INTRODUCTION — The basic requirements of an optimal insulin regimen include administration of a basal insulin plus mealtime boluses of a rapid-acting or short-acting insulin. Basal insulin can be delivered by daily or twice-daily injections of an intermediate-acting (NPH) or long-acting (glargine, detemir, degludec) insulin preparation or by continuous subcutaneous insulin infusion (CSII) via a pump using a rapid-acting insulin preparation (lispro, aspart, glulisine). Providing physiologic insulin replacement requires adjustment of doses to match requirements and relies on glucose monitoring and lifestyle modifications.

This topic will review CSII (insulin pump) therapy. Physiologic insulin replacement, choice of insulin delivery (multiple daily insulin [MDI] injection regimens versus CSII), and designing an MDI regimen are reviewed separately.

●(See "Management of blood glucose in adults with type 1 diabetes mellitus".)

●(See "Overview of the management of type 1 diabetes mellitus in children and adolescents".)

●(See "Insulin therapy for children and adolescents with type 1 diabetes mellitus".)

●(See "Glucose monitoring in the management of nonpregnant adults with diabetes mellitus", section on 'Type 1 diabetes'.)

GENERAL PRINCIPLES — With continuous subcutaneous insulin infusion (CSII), basal insulin is supplied in the form of a continuous infusion (usually comprising between 40 and 50 percent of the total daily insulin dose) with pre-meal bolus doses given to minimize postprandial glucose excursions (see 'Dosing' below). A variety of insulin pumps are available, and the choice among pumps is largely a matter of patient preference, cost, lifestyle, and compatibility with continuous glucose monitoring (CGM) devices.

When using a traditional pump, insulin is infused from a reservoir/cartridge within the pump through tubing to a cannula or needle that is inserted subcutaneously (figure 1). The infusion set (which contains the flexible plastic cannula or steel needle) is secured to the skin by hypoallergenic adhesive. The infusion set and site of infusion are changed by the patient every two to three days. The tubing, which connects the infusion set to the insulin cartridge/reservoir in the pump, can be connected to and disconnected from the infusion site without removing the cannula. The traditional insulin pump can be used with a CGM device as part of a closed-loop system. (See 'Hybrid closed-loop system (artificial pancreas)' below.)

For the patch pump, the insulin reservoir, batteries, and cannula are in a wearable disposable device ("pod"), which delivers insulin subcutaneously (figure 2). The pod is changed every two to three days, and insulin delivery from the pod is controlled wirelessly by a handheld "controller" or personal diabetes manager (PDM). (See 'Types of insulin pumps' below.)

Rapid-acting insulin analogs (lispro, aspart, and glulisine) are typically preferred over regular insulin for continuous insulin therapy. In a meta-analysis of trials comparing rapid-acting insulin analogs with regular insulin for use in CSII, there was a small but significant reduction in glycated hemoglobin (A1C) with use of insulin analogs (mean difference -0.26 percent, 95% CI -0.47 to -0.06 percent) [1]. It was difficult to analyze hypoglycemia as it was defined differently in the various trials. The convenience of being able to administer a rapid-acting insulin immediately before the meal (compared with needing to administer preprandial boluses of regular insulin 30 to 45 minutes before meals), as well as the ability to more quickly correct hyperglycemia, also favor rapid-acting formulations.

For most patients with type 1 diabetes, frequent testing of glucose levels is necessary to achieve A1C targets safely without frequent or severe hypoglycemia. Self-monitoring allows adjustments of doses and timing of insulin as well as the timing and content of meals and snacks based on immediate feedback of glucose results. Many patients with type 1 diabetes use a combination of self-monitoring of blood glucose (SMBG) by fingerstick with a glucose meter and, when available, CGM. Some insulin pumps can receive glucose data from CGM devices, and others use such data to automatically adjust basal rate delivery to address both low and high glucose levels. (See "Management of blood glucose in adults with type 1 diabetes mellitus", section on 'Intensive diabetes therapy' and "Glucose monitoring in the management of nonpregnant adults with diabetes mellitus" and 'Types of insulin pumps' below.)

DOSING

Total daily dose — When converting a patient from a multiple daily insulin (MDI) injection regimen to continuous subcutaneous insulin infusion (CSII), the pre-pump level of chronic glycemia will help determine the initial pump basal rate and preprandial dosing. As an example, for a patient who has been well controlled on the previous MDI regimen (eg, A1C <7 percent), the initial total daily dose (TDD) of insulin administered by pump may be 10 to 20 percent less than the TDD (both short-acting and long-acting insulin) of the previous regimen. Conversely, patients with inadequate glycemic control (high A1C with no hypoglycemia) may be started with the same TDD (both short-acting and long-acting insulin) as they had been using with their injection regimens.

Basal rate — In general, approximately 40 to 50 percent of the TDD is administered as the basal rate (divide by 24 to get units per hour). This proportion can be higher or lower depending on a number of factors, including if the individual consumes a low- or high-carbohydrate diet, respectively. For most patients, basal rates are in the range of 0.01 to 0.015 units per kg per hour (ie, for a 60 kg woman, approximately 0.6 to 0.9 units per hour).

The basal rates are adjusted empirically (eg, by approximately 10 percent) based on glucose monitoring results. The basal rate adjustment depends upon a number of factors, including the degree of hyperglycemia or hypoglycemia, patient fears of hypoglycemia or hyperglycemia, current use of continuous glucose monitoring (CGM) with alarms, age, and presence or absence of comorbidities (eg, more cautious adjustments in older patients with hypoglycemia unawareness or cardiovascular disease).

Certain time periods during the day may require higher, while other periods may require lower infusion rates depending on individual factors including lifestyle and the "dawn phenomenon," which often occurs between 2:00 and 8:00 AM. Most pumps allow for preprogrammed changes in basal rate to accommodate these requirements. The "dawn phenomenon" is thought to result from diurnal secretion patterns of hormones, particularly increased growth hormone at midnight to 2:00 AM, that tend to antagonize the actions of insulin in the early morning hours and so raise blood glucose concentrations. The overnight basal rate(s) can be adjusted to maintain the pre-breakfast blood glucose in the target range [2,3].

Temporary basal rates can also be programmed for defined periods of time. For example, a reduced temporary basal rate can be used to avoid hypoglycemia associated with aerobic activities.

When changing the subcutaneous basal insulin infusion rate, a delay in the actual increase or decrease in plasma insulin levels must be taken into account, based on the kinetics of absorption and time to reach a new steady state [4]. Therefore, basal rates of rapid-acting insulin should be changed approximately two hours before the change in plasma level is required.

Bolus dosing — The pre-meal bolus dose should be based primarily upon the carbohydrate content of the intended meal and the blood glucose level immediately before the meal. If CGM data are available, use of glucose trends (trend arrows) can help patients fine-tune dosing [5]. High fat and protein meals (eg, steak and potatoes or pizza) can cause a prolonged rise in glucose and need for a higher insulin dose and longer duration of insulin action [6,7]. (See "Nutritional considerations in type 1 diabetes mellitus", section on 'Advanced carbohydrate counting'.)

Insulin pumps have insulin calculators for bolus dosing for meals and for correction of hyperglycemia. The use of calculated "active insulin" or "insulin on board" helps prevent hypoglycemia that can occur from overcorrecting for hyperglycemia with multiple correction boluses administered close together in time ("stacking"). Insulin pumps also allow for delivery of extended or dual-wave boluses to help manage the prolonged or delayed rise in glucose concentrations that occur after ingesting higher fat and protein meals or in the presence of gastroparesis.

The timing of dosing preprandially will depend on the insulin used. The kinetics of regular insulin require administration approximately 30 to 45 minutes before eating so that the insulin peak matches the peak glucose after the meal. Rapid-acting insulin can be administered 10 to 15 minutes or immediately before eating, or even during the meal.

FOLLOW-UP VISITS — As in patients with type 1 diabetes treated with multiple daily insulin (MDI) injections, the frequency of clinic visits and adjustments to the insulin regimen vary based on the needs of the patient. This information is reviewed in detail separately. (See "Management of blood glucose in adults with type 1 diabetes mellitus", section on 'Follow-up'.)

TROUBLESHOOTING

Pump failure — Pump failure may occur due to detachment, blockage-kinking, or leakage in the infusion set/cannula, syringe, or connectors, causing an interruption of insulin flow [8,9]. If problems with infusion sets recur, there should be a review of the patient's insertion technique. A different type of infusion set (eg, metal needle instead of Teflon catheter) can be considered. A poor infusion site can also impede insulin delivery. A site may not absorb insulin well due to overuse, scarring, or lipohypertrophy.

There can be pump malfunction due to the need for a battery change. The pump may be in suspension mode, or there may be air in the infusion set causing reduced insulin delivery. Use of expired or damaged insulin (exposure to high temperatures or freezing) will also present as pump failure.

Since the subcutaneous depot is very small and only rapidly acting insulin is administered, any interruption in continuous flow leads very quickly to hypoinsulinemia, and potentially diabetic ketoacidosis. If there is no glucose response to insulin boluses, the infusion set should be immediately changed and a new insertion site chosen. Patients should know how to check for ketones (either blood or urine) and have a back-up plan for both basal and mealtime insulin, which includes insulin syringes or pens, to use in the event of pump failure.

If pump failure occurs, patients can replace basal insulin by taking their rapidly acting insulin every three to four hours (dose of three to four times the hourly basal rate) by injection with careful, frequent blood glucose monitoring. As an alternative, patients can transition to an intermediate-acting or long-acting basal insulin. Using the total daily basal insulin dose as a guide, a similar dose can be given as one injection of insulin glargine or can be divided into two to three injections of NPH daily, or two injections of glargine or detemir. In this setting, we do not use basal insulin with a very long duration of action (eg, degludec), as it takes approximately four days to reach steady state. When restarting pump therapy, if there is residual basal insulin on board, a lower temporary basal rate may be needed until the intermediate-acting or long-acting insulin effect has dissipated.

Superficial infection — Patients who are treated with CSII may rarely develop infections at the site of catheter insertion. This is more likely to occur if the infusion set has not been changed for >72 hours, if the site was not properly prepared, or if the infusion set not properly inserted. If the infusion site is red, swollen, painful, draining purulent material, or has increased warmth, the infusion set should be immediately removed and discarded and a new set placed at another site. The old site should be cleaned and treated with warm soaks; occasionally, antibiotic therapy is needed.

NPO for procedures — Patients who are not eating in preparation for a short procedure may continue with their usual basal infusion rate, assuming that the catheter and pump can remain safely in place during the procedure. Depending on the degree of glucose control before the procedure, the overnight basal rate may be reduced by up to 20 percent to minimize the risk of hypoglycemia the next morning. No boluses would be administered until the patient is able to eat. Catheters may inadvertently be dislodged during procedures, and if the patient is not alert enough to provide self-care, the health care providers should consider changing to injection therapy until the patient is able to manage pump therapy again. (See "Management of diabetes mellitus in hospitalized patients", section on 'Patients with type 1 diabetes'.)

TYPES OF INSULIN PUMPS — A variety of insulin pumps are available, some of which communicate with specific continuous glucose monitoring (CGM) devices. Others use CGM feedback information to automatically adjust basal rate delivery and mitigate both low and high glucose levels. The choice among pumps is largely a matter of patient preference, cost, and lifestyle.

Insulin pumps can be used alone or in conjunction with a CGM device to give the patient more information about their blood glucose levels and allow them to make better informed decisions about insulin dosing. This approach is known as sensor-augmented insulin pump therapy. When the insulin pump uses an algorithm (dependent on CGM results, target glucose, and the amount of active insulin on board) to control basal insulin infusion, the system is termed a partially automated (hybrid) closed-loop system (artificial pancreas). Patients must still determine and administer pre-meal bolus insulin (hence the term, "hybrid"), with dose selection based on insulin-to-carbohydrate ratios and "correction" factors to address ambient glucose levels at the time of the meal. Use of sensor-augmented insulin pump therapy and the hybrid closed-loop system can reduce time in hypoglycemia, but these systems are more costly.

Sensor-augmented insulin pump — A sensor-augmented insulin pump combines the technology of an insulin pump and a CGM device. A randomized trial compared sensor-augmented insulin pump therapy with multiple daily insulin (MDI) injections using standard home blood glucose monitoring (and no CGM) in 329 adults [10]. After one year, the reduction in mean A1C was significantly greater in the pump therapy group (between-group difference -0.6 percentage points). Of note, the design of the study (augmented pump therapy versus MDI without CGM) could not distinguish between the effects of pump therapy and CGM. (See "Glucose monitoring in the management of nonpregnant adults with diabetes mellitus", section on 'CGM systems'.)

Some insulin pumps can be programmed to interrupt insulin delivery (for up to two hours) at a preset sensor glucose value (low glucose threshold suspend). The threshold suspend feature reduces the frequency and duration of nocturnal hypoglycemia, as illustrated by the findings from randomized trials [11,12]:

●In one trial, 247 patients (mean age approximately 43 years) with type 1 diabetes and nocturnal hypoglycemia were randomly assigned to sensor-augmented insulin pump therapy with or without a threshold suspend feature [11]. After three months, nocturnal hypoglycemia (measured as area under the curve) was significantly lower in the group with the threshold suspend feature (1.5 versus 2.2 events per patient per week). Severe hypoglycemia was rare (four episodes), but all events were in control group patients.

●In another trial, 95 patients (mean age 18.6 years) with type 1 diabetes and well-documented hypoglycemia unawareness were randomly assigned to standard insulin pump (without CGM) or to sensor-augmented insulin pump therapy with a threshold suspend feature [12]. After six months, the rate of severe and moderate hypoglycemic events was significantly lower in the group using the threshold suspend feature (9.5 versus 34.2 events per 100 patient-months). Of note, despite randomization, the baseline frequency of moderate and severe hypoglycemia was substantially greater in the threshold suspend group than the control group, which limits interpretation of the results.

In both trials, there was no significant difference in change in A1C, and there were no episodes of diabetic ketoacidosis (ie, no loss of glycemic control with brief suspension of insulin delivery). These findings suggest that the threshold suspend feature is useful, and the first trial shows that it can improve the ability of sensor-augmented insulin pump therapy to reduce nocturnal hypoglycemia [11]. The design of the second trial (augmented pump with threshold suspend feature versus standard pump therapy without CGM) could not distinguish between the effects of CGM and CGM with threshold suspend feature [12].

Newer insulin pumps are available with a "predictive low glucose suspend" feature. In contrast to low glucose threshold suspend, in which insulin delivery is suspended when the glucose reading reaches the threshold value (eg, 70 mg/dL [3.9 mmol/L]), predictive low glucose threshold suspend reduces or suspends insulin infusion when the trend in CGM results predicts that hypoglycemia will occur. In randomized trials of predicative low glucose suspend in children and adults, using different devices, there was a reduction of hypoglycemia without an increase in hyperglycemia [13-15].

Hybrid closed-loop system (artificial pancreas) — Two partially automated (hybrid) closed-loop systems of insulin delivery are commercially available in the United States [16,17]. When using these insulin pump/CGM systems in the "auto" or "automatic" mode, instead of infusing basal insulin in mini-boluses every five minutes according to the programmed basal rates ("manual" mode), the system automatically gives a mini-bolus (or no bolus) of rapidly acting insulin every five minutes determined by an algorithm that is dependent on CGM results, target glucose, and the amount of active insulin on board (figure 1).

With these "hybrid" closed-loop devices, the patient still needs to determine and administer pre-meal insulin boluses, which is facilitated with an individualized insulin-to-carbohydrate ratio set in the pump's bolus calculator. Some systems require periodic fingerstick capillary glucose measurements for calibration and to address high or low values, and some have limited choices for the target glucose. Each of the available devices can transmit insulin dosing data (display of basal and bolus insulin delivery), CGM and SMBG data, as well as pump and CGM settings to cloud-based systems. These data can be retrieved and reviewed on demand (figure 3A-B and figure 4A-B).

For the first commercial hybrid closed-loop system in the United States, the few available small reports indicate that discontinuation rates in the real world are high [18,19]. Substantial education and support for the patient as well as considerable diligence by the patient regarding self-care tasks are required to stay in "auto mode"; improvements are anticipated in future models.

In a meta-analysis of trials comparing the use of any hybrid closed-loop system with any insulin-based treatment in nonpregnant patients with type 1 diabetes, the proportion of time spent near normoglycemia (70 to 180 mg/dL [3.9 to 10 mmol/L]) over 24 hours was modestly, albeit significantly, higher with the hybrid closed-loop system (weighted mean difference 9.62 percent, 95% CI 7.54-11 percent) [20]. Overall, the incidence of severe hypoglycemia was low in both groups. Most of the trials examined short-term (one to three days) control [21-25]. Only a few of the trials examined the utility of these devices in the outpatient setting, during eating and usual daily activities, over a longer period [16,26-32]. As examples:

●In a crossover, random-order trial, 33 adults (mean A1C 8.5 percent [69.4 mmol/mol]) were assigned to either 12 weeks of partially automated (hybrid), closed-loop insulin delivery (intervention) followed by 12 weeks of sensor-augmented pump therapy (control), or to the opposite order (sensor-augmented pump therapy followed by hybrid, closed-loop insulin delivery) [27]. Patients performed their usual daily activities and were not monitored remotely by study staff.

Compared with the sensor-augmented pump, use of the hybrid closed-loop system resulted in a greater proportion of time spent in the target range of 70 to 180 mg/dL (3.9 to 10 mmol/L; 67.7 versus 56.8 percent, mean difference 11 percentage points, 95% CI 8.1-13.8). The mean glucose level (157 versus 168 mg/dL) and the mean A1C level (7.3 versus 7.6 percent) were also lower during the closed-loop phase of insulin delivery. Hypoglycemia, as measured by the area under the curve when glucose was <63 md/dL (3.5 mmol/L), was lower during the closed-loop system than during the control period (169 versus 198 [mg/dL x min]).

●In a trial in children and adolescents with the same device, but delivering insulin only overnight, 25 patients (mean A1C 8.1 percent [65 mmol/mol]) used the hybrid closed-loop insulin delivery system overnight and discontinued it before breakfast [27]. Compared with the sensor-augmented insulin pump, use of the hybrid closed-loop system resulted in a greater proportion of nocturnal time spent with glucose levels in the target range of 70 to 145 mg/dL (3.9 to 8 mmol/L; 59.7 versus 34.4 percent, mean difference 24.7 percentage points, 95% CI 20.6-28.7). The mean overnight glucose level was lower with the closed-loop system (146 versus 176 mg/dL). The proportion of time spent with a blood glucose level <70 or <50 mg/dL (<3.9 or <2.8 mmol/L) was low and was not reduced during the closed-loop treatment arm (<4 and <1 percent, respectively).

In a subsequent six-month trial comparing a hybrid closed-loop system with a sensor-augmented insulin pump in 168 patients ≥14 years of age, the percentage of time in target range (70 to 180 mg/dL [3.9 to 10 mmol/L]) as measured with CGM was higher in the closed-loop group (71 versus 59 percent, risk-adjusted difference 11 percent, 95% CI 9-14) [17]. A1C levels improved in patients using the closed-loop system (7.4 to 7.06 percent) but did not change in controls (7.4 to 7.39 percent). Although there were no serious hypoglycemic events in either group, the percentage of time spent in hypoglycemia was lower in patients assigned to the closed-loop system (eg, <54 mg/dL, 0.29 versus 0.35 percent, risk-adjusted difference -0.10, 95% CI -0.19 to -0.02). There were, however, more hyperglycemic adverse reactions, including one episode of ketoacidosis, in the closed-loop group (14 versus 2 patients), primarily due to infusion set failures. In a similarly designed 16-week trial in children 6 to 13 years of age, the percentage of time in target range was higher with the closed-loop system (67 versus 55 percent, mean adjusted difference 11 percentage points, 95% CI 7-14) [33].

Bihormonal, completely automated system — Studies have evaluated the efficacy of a fully automated closed-loop system of insulin delivery based upon continuous glucose sensing (artificial pancreas). The automated, bihormonal, closed-loop system uses two commercially available pumps, with one delivering insulin and the other glucagon. This bihormonal, completely automated system is not commercially available.

●In a crossover trial, 43 adults received therapy with automated, bihormonal (insulin and glucagon), closed-loop system for 11 days and therapy with their own insulin pump for 11 days [31]. The bihormonal pump therapy only requires input of the patient's body mass index (BMI) to initiate therapy. The delivery of the insulin and glucagon during the closed-loop arm was determined completely automatically by an algorithm that was, in turn, dependent on CGM results. During the closed-loop system part of the trial, patients continued all normal activities, including exercise and driving. Fingerstick plasma glucoses were checked at least four times daily; however, the closed-loop system was driven entirely by the CGM using commercially available devices.

The mean CGM glucose concentration was lower during the closed-loop system, as compared with the comparator period (140.4 versus 162 mg/dL [7.8 versus 9.0 mmol/L]), and the percentage time with a glucose level <60 mg/dL (3.3 mmol/L) was lower (0.6 versus 1.9 percent, respectively). There were no severe hypoglycemic events during the closed-loop period.

●In a similarly designed, five-day crossover trial involving 32 adolescents, patients participated in the same activities, ate the same meals, and stayed in the same cabins as nonparticipants at a diabetes camp [26]. On days 2 through 5 of the closed-loop system, as compared with the control period, the mean glucose level was lower (142 versus 158 mg/dL [7.0 versus 8.8 mmol/L]) and the percentage time with a glucose level <70 mg/dL (3.9 mmol/L) or <60 mg/dL (3.3 mmol/L) was similar (3.1 and 4.9 percent and 1.3 and 2.2 percent, respectively). There were no severe hypoglycemic events during the closed-loop period.

Do-it-yourself artificial pancreas systems (DIY APS) — The use of do-it-yourself artificial pancreas systems (DIY APS) or "looping" by individuals with type 1 diabetes to automatically infuse insulin has increased in the United States and globally. The systems use commercially available insulin pumps and CGM devices, smart phones, and experimental applications that control insulin infusion. Additional hardware and communication devices may be needed. Free open-source resources for each of the available DIY looping systems are offered on their websites. None of these systems are US Food and Drug Administration (FDA) approved, all are considered experimental, and all involve a high degree of patient involvement.

SOCIETY GUIDELINE LINKS — Links to society and government-sponsored guidelines from selected countries and regions around the world are provided separately. (See "Society guideline links: Diabetes mellitus in adults" and "Society guideline links: Blood glucose monitoring".)

SUMMARY AND RECOMMENDATIONS

●With continuous subcutaneous insulin infusion (CSII; insulin pump), basal insulin is supplied in the form of a continuous infusion (usually comprising between 40 and 50 percent of the total daily dose [TDD]) with pre-meal bolus doses given to minimize postprandial glucose excursions. (See 'General principles' above.)

●We suggest rapid-acting insulin analogs (lispro, aspart, and glulisine) rather than regular insulin for continuous insulin therapy (Grade 2C). (See 'General principles' above.)

●When converting a previously well-controlled patient from a multiple daily insulin (MDI) injection regimen to CSII, the initial TDD of insulin administered by pump may be 10 to 20 percent less than the TDD (both short-acting and long-acting insulin) of the previous regimen. Conversely, patients with inadequate glycemic control (high glycated hemoglobin [A1C] with no hypoglycemia) may be started with the same TDD (both short-acting and long-acting insulin) as they had been using with their injection regimens. (See 'Total daily dose' above.)

●Approximately 40 to 50 percent of the TDD is administered as the basal rate (divide by 24 to get units per hour). This proportion can be higher or lower depending on a number of factors, including if the individual consumes a low- or high-carbohydrate diet, respectively. For most patients, basal rates are in the range of 0.01 to 0.015 units per kg per hour (ie, for a 60 kg woman approximately 0.6 to 0.9 units per hour). (See 'Basal rate' above.)

●The pre-meal bolus dose should be based primarily upon the carbohydrate content of the intended meal and the blood glucose level immediately before the meal. Many insulin pumps have insulin calculators for bolus dosing for meals and for correction of hyperglycemia. Extended or dual-wave boluses can be used to help manage the prolonged or delayed rise in glucose that is commonly observed after eating higher fat and protein meals or in the presence of gastroparesis. The use of calculated "active insulin" or "insulin on board" helps prevent hypoglycemia that can occur from overcorrecting for hyperglycemia with multiple correction boluses ("stacking"). (See 'Bolus dosing' above.)

●The frequency of clinic visits and adjustments to the insulin regimen vary based on the needs of the patient. (See "Management of blood glucose in adults with type 1 diabetes mellitus", section on 'Follow-up'.)

●A variety of insulin pumps are available, and the choice among pumps is largely a matter of patient preference, cost, lifestyle, and compatibility with continuous glucose monitoring (CGM) devices (figure 2 and figure 1). Use of sensor-augmented insulin pump therapy and the hybrid closed-loop system can reduce time in hypoglycemia, but these systems are more costly. (See 'Types of insulin pumps' above.)