Prevalence of Insulin Pump Therapy and Its Association with Measures of Glycemic Control: Results from the Canadian Study of Longevity in Type 1 Diabetes

Abstract

Objective:

We aimed to determine cross-sectional insulin pump prevalence and factors associated with measures of glycemic control as a secondary analysis in a long-standing type 1 diabetes mellitus (T1DM) national cohort.

Research Design and Methods:

Canadian participants with ≥50 years of T1DM (n = 305) were administered a comprehensive mail-based questionnaire including acquisition of contemporaneous laboratory results. Factors associated with pump use, glycosylated hemoglobin (HbA1c), and hypoglycemia were analyzed by regression.

Results:

The 305 participants had a median age of 65 [interquartile range, 59, 71] years, median diabetes duration of 54 [51, 59] years, and mean HbA1c level of 7.5 ± 1.1%. Prevalence of pump use was 44% (133/305), with median duration of use 8 [4, 13] years. Compared with the non-pump subgroup, the pump subgroup had numerically lower but similar HbA1c levels (7.4 ± 0.9% vs. 7.6 ± 1.2%; P = 0.22) and reported greater numbers of minor hypoglycemia events (6.5 vs. 5.1 events/patient·month; P = 0.004) and fewer severe hypoglycemia events (0.5 vs. 1.3 events/patient·year; P = 0.02) in the past year. More frequent daily glucose tests and more frequent minor hypoglycemia events—but not pump therapy or its prescription parameters—were independently associated with lower HbA1c level in multivariable regression. However, use of insulin pump and habitual use of continuous glucose monitoring (≥1 week/month) were each independently associated with lower risk of severe hypoglycemia (risk ratio = 0.50 [P < 0.0001] and 0.30 [P = 0.001], respectively).

Conclusions:

Insulin pump and continuous glucose monitoring technologies were associated with lower risk of severe hypoglycemia, while frequent daily glucose testing was associated with lower HbA1c level. These findings imply that basic self-management skill and technology play complementary roles in glycemic control among older adults with long-standing T1DM.

Introduction

Parallel to the systematic implementation of intensive insulin therapy for type 1 diabetes mellitus (T1DM) management since the initial publication of the Diabetes Control and Complication Trial¹ over two decades ago, the burden of micro- and macrovascular complications has decreased substantially but incompletely.² As the key predictors of diabetes complications are the duration of diabetes and the intensity of hyperglycemic exposure, patients with a diagnosis of diabetes that predates this era of intensive glycemic control likely remain at high cumulative risk of complications.^1,3,4

As a key intervention against hyperglycemic exposure and complications risk, insulin pump therapy has become a more common therapeutic option in the last two decades and was reported being used by approximately 10% of patients with T1DM in Canada in 2009.⁵ This greater uptake of insulin pump therapy has likely occurred as a result of the perception that such use of continuous subcutaneous insulin delivery makes treatment goals further achievable by way of more physiological insulin delivery. Systematic reviews, of which the most contemporary analysis included 23 randomized trials involving 976 participants with T1DM randomized to insulin pump therapy or multiple daily injections, demonstrated a small but significant mean difference of glycosylated hemoglobin (HbA1c) of −0.25% in favor of the insulin pump.^6

–9 Further benefits are less clear but may include a reduction of total insulin dose, incidence of severe hypoglycemia episodes, and glycemic variability.^9
–11 Furthermore, it is commonly understood that the ability to program more physiologically appropriate basal rate increments according to the time of day and certain convenience factors, such as the calculation of food and correction bolus insulin, may assist in more effective glycemic control compared with multiple daily injection therapy.¹²

In older patients with T1DM diabetes in whom the risk of hypoglycemia is amplified,¹³ the prevalence of insulin pump use and its possible beneficial effect on glycemic control are not known as studies have mainly assessed cohorts of young participants with shorter diabetes duration.^14,15 The Canadian Study of Longevity in Type 1 Diabetes was initiated in 2013 and identified Canadians living with diabetes for at least 50 years. The primary objective of this cohort study is to determine factors associated with complications and the biomarkers for their early stages. However, this cohort represents a unique opportunity to explore in secondary analysis the nature and impact of insulin pump therapy in an older population. We aimed to determine the prevalence of insulin pump use, the factors associated with its use (with a particular focus on insulin pump parameters such as basal rate profiles and the quantitative use of food and correction boluses), and its independent association with glycemic control as measured by HbA1c and reported severe hypoglycemia events.

Research Design and Methods

The Canadian Study of Longevity in Type 1 Diabetes is a cross-sectional study of Canadians living with T1DM for 50 years or more, with the primary objective to establish a baseline population registry to determine the factors associated with resistance to the development of complications after long-standing disease duration.^16,17 The current report represents a secondary exploratory analysis. Between April 2013 and December 2014, 427 people across Canada responded to public advertisements, online media, and mailings to healthcare professionals, with the support of the Canadian Diabetes Association, the JDRF Canada, and its Canadian Clinical Trial Network. Enrollment criteria included a documented history of T1DM for 50 years or more and the ability to understand and cooperate with the study questionnaire procedures. Our criterion for duration of diabetes was defined according to previous longevity studies in T1DM.^18,19 Respondents received a detailed 35-page questionnaire by mail, and in five cases the respondent was interviewed by phone. Of the 427 respondents, 386 agreed to participate and received a questionnaire. Nineteen withdrew consent after the mailing. Of the 367 active participants, 305 completed the questionnaire by the time of data lock for this analysis, which occurred because the anticipated sample size of 300 was reached. The Research Ethics Board of the participating institution, Mount Sinai Hospital (Toronto, ON, Canada), approved the protocol, and participants provided written informed consent.

Data collection

In our mail-based questionnaire, participants were extensively surveyed about their diabetes and its management, as well as their past and present medical history. We obtained their latest clinical laboratory report, including measures such as cholesterol levels, HbA1c level, and estimated glomerular filtration rate according to the Modification of Diet in Renal Disease Study equation (eGFR_MDRD).²⁰ In addition, results from the most recent funduscopic examination were obtained directly from their healthcare providers.

Insulin administration parameters

Participants underwent detailed assessment of current insulin therapy and were dichotomized into two subgroups: the non-pump subgroup and the pump subgroup. Both subgroups were questioned on the use and frequency of continuous glucose monitoring (CGM), average number of daily capillary glucose tests, data uploads from meters or insulin pumps, and the type and total daily dose of their insulin, including quantification of basal and bolus insulin.

CGM use was grouped into three categories: those having never adopted CGM use (accounting for 266 [87%] participants), those with nonhabitual use of CGM (defined as less than 1 week per month, accounting for 16 [5%] participants), or those with habitual use of CGM (≥1 weeks per month, accounting for 23 [8%] participants).

The types of insulin were categorized as “analog” if the brands used were insulin detemir, insulin glargine, insulin aspart, insulin glulisine, or insulin lispro. Participants were questioned about use of correction doses for hyperglycemia, as well as on their food bolus dosing strategy according to fixed dose prescription (“Fixed Dose”), use of an estimation range depending on meal size (“Estimation Range”), or use of a quantitative method of carbohydrate counting using a carbohydrate ratio (“Carb Counting”). Multiple responses from the same participant were handled by assigning the method requiring the lowest degree of ability (Fixed Dose was considered lowest, Carb Counting the highest).

In the pump subgroup, the presence of an insulin basal profile adjusted for the dawn phenomenon was defined by an hourly insulin infusion rate that was at least 10% higher during a sampled overnight time segment (3:00 a.m. and 6:00 a.m.) compared with the average hourly infusion rate for the remaining hours of the day.¹²

Lifestyle factors

Adherence to physical activity was defined qualitatively as the report of current regular exercise. Dietary adherence was defined by the presence of two criteria: the reported attempt to limit excess intake of carbohydrate and fat and the regular consumption of fruits and vegetables. Quality of life measures included (1) participant's self-assessment of his or her current quality of life, with possible responses of excellent, good–normal, or poor; (2) assessment of depression by way of the validated Geriatric Depression Scale, with scores ranging from 0 to 15 and for which higher scores indicated greater quantitative depressive symptoms^21,22; and (3) diabetes-related emotional distress by way of the Problem Areas In Diabetes questionnaire, with scores ranging from 0 to 100 and for which higher scores indicated greater distress as determined from 20 questions on dietary, emotional, and social support.²³

Diabetes complications

Participants were asked subjectively about history of micro- and macrovascular complications of diabetes. Objective data were also obtained as follows: (1) Objective diabetic neuropathy was defined by a score of 3 or more on the 15-item questionnaire section of the Michigan Neuropathy Screening Instrument.^24,25 (2) Objective diabetic nephropathy was defined by a contemporary albumin-to-creatinine ratio of ≥2 mg/mmol if a medication of renin–angiotensin system inhibitor was used,²⁶ or 3.3 mg/mmol without,²⁷ or an age-adjusted estimated glomerular filtration rate (eGFR_MDRD) lower than 60 mL/min.²⁶ (3) Objective retinopathy was based on contemporary funduscopic examination results from the participant's clinical ophthalmologist's or optometrist's report.

Minor and severe hypoglycemia

Self-reported estimates of minor and severe hypoglycemia were obtained by questionnaire. Minor hypoglycemia was defined by self-reported capillary blood sugar reading of <4 mmol/L that did not require help from others, whereas severe hypoglycemia was defined as events that required help from a friend or family member or required emergency medical services or a visit to a hospital. Patients were questioned on lifelong events and, to reduce recall bias, the frequency of events in the past calendar year. Assessment of fear of hypoglycemia was determined by the Hypoglycemia Fear Questionnaire, with scores ranging from 0 to 132 and for which higher scores indicated greater magnitude of fear according to participant's behaviors and worries related to hypoglycemia.²⁸

Statistics

SAS (version 9.4) software (SAS Institute Inc., Cary, NC) was used for all statistical analyses. Sample size calculations were based on 1962 census date and a contemporaneous, age-specific incidence rate of T1DM.^29,30 From this we estimated that approximately 1,441 Canadians would be eligible for analysis in 2013. We estimated a conservative 20% response based on current cohort experience³¹ and aimed for a total sample size of approximately 300 participants, which provided 86.9% power to detect a 0.25% difference in HbA1c between insulin pump users and injections users, under the assumption of a conservative SD of 0.70%.⁹ Variables were calculated as mean ± SD values or as the median and interquartile range, if the value was not normally distributed.

The general characteristics and the technical factors related to diabetes care were compared according to pump and non-pump subgroup membership, with Student's t test or the Wilcoxon rank-sum test for continuous variables and χ² test for categorical variables. Sensitivity analysis was planned in which the statistical procedures were repeated on the cohort in whom those with two or fewer injections per day were excluded. Linear regression analysis was performed to assess the relationship between HbA1c and the general characteristics and technical factors related to diabetes care in univariable analyses. Multivariable linear regression was performed with the variables significantly associated with HbA1c in univariable analysis as the independent variables; age, insulin pump use, CGM use, and hypoglycemia events were included regardless of significance. As a sensitivity analysis, the multivariable regression was also run using a backward elimination variable selection model, with the same model structure as above. Linear regression analysis was performed using the frequency of minor hypoglycemia events in the past calendar year as the outcome variable, using the same univariable screening process for a final multivariable model. In addition, Poisson regression analysis was performed using the frequency of severe hypoglycemia events in the past calendar year as the outcome variable, using the same univariable screening process for a final multivariable model.

P values of <0.05 were considered statistically significant.

Results

The 305 participants were 45% male and had a median age of 65 [59, 71] years, median diabetes duration of 54 [51, 59] years, mean body mass index of 26 ± 4.7 kg/m², and mean HbA1c of 7.5 ± 1.1% (Table 1). One hundred and thirty-three subjects (44%) were on insulin pump therapy.

Table 1.

General Characteristics of the Groups According to Insulin Delivery Method

	Total	Non-pump subgroup	Pump subgroup	P value ^a
Participants, n (%)	305	172 (56%)	133 (44%)
Age (years)	65 [59,71]	66 [60,73]	64 [59,69]	0.04
Male, n (%)	137 (45%)	82 (48%)	55 (42%)	0.30
Age (years) at diagnosis	11.3 ± 7.3	11.9 ± 8.0	10.5 ± 6.2	0.10
Duration (years) of diabetes	54 [51, 59]	54 [51, 60]	53 [51, 57]	0.12
Body mass index (kg/m²)	26.0 ± 4.7	26.2 ± 5.1	25.7 ± 4.2	0.41
Level of education higher than high school, n (%)	239 (78%)	126 (73%)	113 (85%)	0.01
Mean systolic blood pressure (mm Hg)	128.9 ± 15.0	129.7 ± 15.2	127.7 ± 14.9	0.41
HDL cholesterol (mmol/L)	1.72 ± 0.52	1.66 ± 0.50	1.79 ± 0.54	0.05
LDL cholesterol (mmol/L)	2.02 ± 0.69	2.00 ± 0.71	2.06 ± 0.68	0.49
Triglycerides (mmol/L)	0.92 ± 0.53	0.99 ± 0.63	0.83 ± 0.37	0.02
HbA1c (%)	7.5 ± 1.1	7.6 ± 1.2	7.4 ± 0.9	0.22^b
Diabetes complications, n (%)
Diabetic retinopathy^c	192 (72%)	104 (70%)	88 (73%)	0.58
Neuropathy (MNSI questionnaire ≥3)	128 (42%)	74 (43%)	54 (41%)	0.68
Nephropathy (ACR ≥2 mg/mmol)	89 (38%)	60 (45%)	29 (29%)	0.01
Nephropathy (GFR <60 mL/min/1.73 m²)	95 (33%)	56 (34%)	39 (31%)	0.57
Macrovascular complications^d	105 (34%)	66 (38%)	39 (29%)	0.10
Quality of life
QOL self-assessment, n (%)
Excellent	118 (39%)	60 (35%)	58 (44%)	0.07
Good–normal	178 (59%)	107 (62%)	71 (54%)
Poor	8 (3%)	5 (3%)	3 (2%)
Problem Areas in Diabetes score (/100)	10.0 [5.0, 18.0]	10.0 [5.0, 19.0]	9.0 [4.0, 17.5]	0.27
Geriatric Depression Scale score (/15)	1.0 [0.0, 3.0]	2.0 [0.0, 4.0]	1.0 [1.0, 3.0]	0.65

Data are mean ± SD values, median [interquartile range], or n (%). Percentages are calculated from available data.

P values for comparison between non-pump and pump subgroups were calculated using Student's t test, the Wilcoxon rank-sum test, or the χ² test.

Glycosylated hemoglobin (HbA1c) between the subgroups after exclusion of the non-pump users with insulin injections twice daily or less was also not significantly different (HbA1c, 7.4 ± 0.9% vs. 7.5 ± 1.2%; P = 0.50).

Presence of any degree of diabetic retinopathy as determined by most recent funduscopic examination results.

Macrovascular complications as determined by history of myocardial infarction, angina, cardiac or leg artery bypass surgery, or angioplasty.

ACR, albumin-to-creatinine ratio; GFR, glomerular filtration rate; HDL, high-density lipoprotein cholesterol; LDL, low-density lipoprotein; MNSI, Michigan Neuropathy Screening Instrument; QOL, quality of life.

The non-pump subgroup consisted of 172 individuals who did not report current insulin pump use. Of these, none reported no injections, the majority reported three or more insulin injections per day, and 13 reported two insulin injections per day. Of these 13 individuals, two reported use of premixed insulin, whereas the remainder used a basal and a single bolus insulin dose per day.

The pump subgroup consisted of 133 participants. The median duration of pump therapy was 8 [4, 13] years, with seven participants treated for less than 2 years and 20 participants for 20 years or more. Compared with the non-pump subgroup (n = 172, 56%), the pump subgroup was slightly younger (median, 64 [59, 69] vs. 66 [60, 73] years; P = 0.04), was more educated (85.0% vs. 73.3% with an education level higher than high school; P = 0.01), and had lower triglyceride levels (0.83 ± 0.37 vs. 0.99 ± 0.63 mmol/L; P = 0.02). Although the mean HbA1c was 0.2% lower in the insulin pump subgroup, this difference was not significantly different from the non-pump subgroup (7.4 ± 0.9% vs. 7.6 ± 1.2%; P = 0.22). No significant difference was found between the two treatment groups in regard to gender, duration of diabetes, body mass index, and micro- and macrovascular complications, except for a lower prevalence of microalbuminuria in the pump subgroup (28.7% vs. 45.1%; P = 0.01). Finally, results of the Problem Areas in Diabetes score and of the Geriatric Depression Scale score did not reveal differences in diabetes-related distress or differences in mood between the subgroups.

Table 2 shows the technical factors related to diabetes care for the 305 participants, also presented according to pump and non-pump subgroups. Compared with the non-pump subgroup, the pump subgroup was more likely to be physically active (77.3% vs. 64.5%; P = 0.02), monitor capillary glucose tests more frequently (daily glucose tests, 6 [4.0, 7.0] vs. 4 [4.0, 6.0]; P < 0.0001), upload blood glucose data on a computer (38% vs. 11%; P < 0.0001), use CGM (habitual use, 11% vs. 5%; nonhabitual use, 11% vs. 1%; P = 0.0003), and attend follow-up visits with an endocrinologist (86.1% vs. 74.5%; P = 0.02).

Table 2.

Technical Factors Related to Diabetes Care According to Insulin Delivery Method

	Total (n = 305)	Non-pump subgroup (n = 172)	Pump subgroup (n = 133)	P value ^a
Habits
Healthy diet	164 (54%)	93 (54%)	71 (53%)	0.91
Currently physically active	214 (70%)	111 (65%)	102 (77%)	0.02^b
Self-assessment
Daily frequency of capillary glucose tests	5.0 [4.0. 6.5]	4.0 [4.0. 6.0]	6.0 [4.0, 7.0]	<0.0001^b
Upload blood glucose data	68 (24%)	18 (11%)	50 (38%)	<0.0001^b
Use of CGM^c
Nonhabitual	16 (5%)	2 (1%)	14 (11%)	0.0003
Habitual	23 (8%)	8 (5%)	15 (11%)
Medical care
Follow-up by endocrinologist	225 (80%)	114 (75%)	111 (86%)	0.02^b
Medical visits ≥2/year	255 (84%)	139 (81%)	116 (87%)	0.10
Insulin therapy
Daily insulin dose (units/kg)	0.52 ± 0.24	0.56 ± 0.27	0.47 ± 0.16	0.01^b
Number of insulin injections per day	—	4.0 [4.0, 5.0]	—	—
Basal insulin
Number of units per day (units)	16.0 [12.0, 23.0]	18.0 [13.0, 28.0]	14.9 [10.8, 20.0]	0.001^b
Percentage of total daily dose	50 [43, 61]	53 [44, 65]	48 [40, 55]	0.004^b
Dawn phenomenon coverage (%)	—	—	36.2	—
Use of analog insulin	275 (92%)	146 (86%)	129 (99%)	<0.0001^b
Bolus insulin
Number of units per day (units)	15.9 [11.1, 22.7]	16.0 [11.0, 24.0]	15.8 [11.4, 22.3]	0.72
Percentage of total daily dose	50 [39, 58]	47 [35, 56]	52 [45, 60]	0.01^b
Use of analog insulin	281 (94%)	152 (90%)	129 (99%)	0.001^b
Food bolus				<0.0001^b
Fixed Dose	38 (12%)	30 (17%)	7 (5%)
Estimation Range	166 (54%)	122 (71%)	44 (33%)
Carb Counting	102 (33%)	20 (12%)	82 (62%)
Correction bolus use	265 (89%)	144 (85%)	121 (95%)	0.004^b

Data are mean ± SD values, median [interquartile range], or n (%). Percentages are calculated from available data.

P values for comparison between non-pump and pump subgroups were calculated using Student's t test, the Wilcoxon rank-sum test, or the χ² test.

Statistically significant difference, P < 0.05.

Nonhabitual continuous glucose monitoring (CGM) use was defined as <1 week per month, and habitual CGM use was defined as ≥1 weeks per month. The remaining 266 (87%) participants never adopted CGM use.

Carb, carbohydrate.

The daily insulin dose was significantly lower in the pump subgroup (0.47 ± 0.16 vs. 0.56 ± 0.27 units/kg/day; P = 0.01). The non-pump subgroup administered a median of 4 [4, 5] injections per day, while 13 (7.6%) of them had two or fewer injections per day. From the total daily dose of insulin, the proportion of basal insulin was a median of 48.0% [40.3, 54.9] in the pump subgroup, significantly lower compared with 52.6% [44.1, 65.3] in the non-pump subgroup (P = 0.004). Among the pump subgroup, 36.2% had basal insulin rate programming consistent with the “dawn phenomenon,” and compared with the non-pump subgroup they were more likely to use analog insulin (99.2% vs. 86.4%; P < 0.0001). In the pump subgroup, regular human insulin was used by one participant (0.8%). Moreover, the pump subgroup used a higher proportion of bolus insulin (52.0% [45.1, 59.7] vs. 47.4% [34.7, 55.9]; P = 0.01) and demonstrated greater use of analog insulin (99.2% vs. 89.9%; P = 0.001) and of carbohydrate ratios for the calculation of meal boluses (61.7% vs. 11.6%; P < 0.0001). The majority of both groups reported using a correction dose for hyperglycemia, but it was significantly more prevalent in the insulin pump group (95.3% vs. 84.7%; P = 0.004).

Self-reported estimates of minor and severe hypoglycemia according to insulin delivery are summarized in Table 3. Minor hypoglycemia was slightly, but significantly, greater in the pump subgroup, in which participants reported a mean of 6.5 compared with 5.1 events per month (P = 0.004). However, despite a similar proportion of patients who experienced severe hypoglycemia over their lifetime, substantially fewer patients in the pump subgroup reported a severe hypoglycemia event in the past year (20% in the pump subgroup compared with 32% in the non-pump subgroup; P = 0.02). Although hypoglycemia unawareness was common in the entire group (41%), it did not differ according to the subgroups. The level of fear of hypoglycemia did not differ between subgroups, either in the total integrated score or in its Behavior and Worry Subscales.

Table 3.

Self-Reported Minor and Severe Hypoglycemia Events According to Insulin Delivery Method

	Total (n = 305)	Non-pump subgroup (n = 172)	Pump subgroup (n = 133)	P value
Minor hypoglycemia^a
Events in past calendar year (n per patient-month)	5.7 ± 9.0	5.1 ± 9.9	6.5 ± 7.6	0.004
Severe hypoglycemia^b
Lifetime (% of patients)	234 (77%)	132 (78%)	102 (77%)	0.84
Past calendar year (% of patients)	76 (27%)	51 (32%)	25 (20%)	0.02
Events in past calendar year (n per patient-year)	1.0 ± 2.9	1.3 ± 3.7	0.5 ± 1.4	0.02
Hypoglycemia unawareness (% of patients)^c	122 (41%)	71 (43%)	51 (39%)	0.54
Hypoglycemia Fear Questionnaire^d
Total Score	28 [16, 43]	31 [17, 44]	26 [15, 41]	0.14
Behavior Subscale	13 [9, 21]	15 [9, 21]	13 [8, 19]	0.20
Worry Subscale	13 [6, 24]	14 [7, 25]	11 [6, 23]	0.24

Data are mean ± SD values, median [interquartile range], or n (%). Percentages are calculated from available data.

Minor hypoglycemia defined by self-reported capillary blood sugar reading of less than 4 mmol/L that did not require help from others.

Severe hypoglycemia defined as events that required help from a friend or family member or required emergency medical services or a visit to a hospital.

Hypoglycemia unawareness was defined by the qualitative presence of lost ability to feel the symptoms of hypoglycemia.

Hypoglycemia Fear Questionnaire scores could range from 0 to 132, with higher scores indicating greater magnitude of fear, as a composite of behaviors and worries related to hypoglycemia.²⁸

Age, gender, and the general characteristics and technical factors related to diabetes care significantly associated with HbA1c in univariable regression are shown in Table 4, under the univariable analysis column headings. In this analysis, lower HbA1c level was associated with male gender, lower low-density lipoprotein cholesterol level, lower triglyceride level, lower Problem Areas in Diabetes score (representative of less diabetes distress), lower Geriatric Depression Scale score (representative of fewer depressive symptoms), and fewer macrovascular complications. Furthermore, lower HbA1c level was associated with presence of regular exercise, higher frequency of daily capillary glucose tests, and higher percentage of daily bolus insulin. In the multivariable analysis, the variables independently associated with lower HbA1c were male gender, higher frequency of daily capillary glucose tests, and greater frequency of minor hypoglycemia (without an association with severe hypoglycemia). Insulin pump use was associated with HbA1c in neither univariable nor multivariable models. The results of the backward elimination variable selection approach were similar to those of the stepwise process presented (Table 4).

Table 4.

Univariable and Multivariable Linear Regression Analysis of Variables Associated with Glycosylated Hemoglobin in the 305 Participants

	Univariable analysis ^a		Multivariable analysis ^b
Characteristic	β-Coefficient	P value	β-Coefficient	P value
General
Age	−0.01	0.30	−0.01	0.24
Gender (reference female)	−0.56	<0.0001^c	−0.44	0.0046^c
LDL cholesterol	0.25	0.01^c	0.08	0.43
Triglycerides	0.30	0.02^c	0.19	0.26
PAID score	0.01	0.003^c	0.01	0.33
Geriatric Depression Scale score	0.06	0.02^c	0.03	0.55
Macrovascular complications	0.28	0.03^c	0.25	0.12
Habits
Currently physically active	−0.28	0.04^c	0.05	0.78
Self-assessment
Daily frequency of capillary glucose tests	−0.12	<0.0001^c	−0.09	0.01^c
CGM use^d
Nonhabitual	0.18	0.81	0.32	0.52
Habitual	−0.04		0.16
Hypoglycemia (number of events in past calendar year)
Minor (per month)	−0.03	<0.0001^c	−0.02	0.0052^c
Severe	0.01	0.58	0.0001	0.97
Insulin therapy
Percentage of daily bolus insulin	−0.01	0.03^c	0.001	0.83
Use of an insulin pump	−0.15	0.24	−0.04	0.81

All variables from Tables 1 –3 were analyzed in univariable linear regression, but only the significant results are showed here, with exceptions for continuous glucose monitoring (CGM) use, the number of severe hypoglycemia events, and the use of an insulin pump.

All significant variables from univariable linear regression (P < 0.25) as well as use of CGM and the number of severe hypoglycemia events in the past year were included in the multivariable analysis, and age was forced into the model (data not shown). R ² value = 0.26. In a second analysis strategy for multivariable analysis (data not shown), using a backward elimination variable selection model, male gender, higher frequency of daily capillary glucose test, and number of minor hypoglycemia events in remained significantly associated with lower glycosylated hemoglobin level, along with a lower Problem Areas in Diabetes (PAID) score.

Statistically significant difference, P < 0.05.

Nonhabitual CGM use was defined as <1 week per month, and habitual CGM use was defined as ≥1 weeks per month. The reference group consisted of participants who never adopted CGM use.

LDL, low-density lipoprotein.

In univariable linear regression analysis using frequency of minor hypoglycemia events in the past calendar year as the outcome variable, lower HbA1c, lower triglyceride level, higher frequency of daily capillary glucose tests, lack of presence of macrovascular complications, and current physical activity were associated with more frequent minor hypoglycemia. In the multivariable model, which included age as a covariate, HbA1c and frequency of daily capillary glucose tests remained associated with minor hypoglycemia (data not shown). Insulin pump use was not associated with minor hypoglycemia in univariable or multivariable models. However, we present in Table 5 the factors associated with severe hypoglycemia. Columns 2 and 3 show the univariable risk ratios (and corresponding P values), and the multivariable model is shown in columns 4 and 5. We note that age, presence of macrovascular complications, habitual CGM use, minor hypoglycemia, Hypoglycemia Fear Questionnaire score, and use of an insulin pump were all associated with severe hypoglycemia in the multivariable model. This indicates that insulin pump use and the habitual use of CGM were each independently associated with a substantially lower risk of reported severe hypoglycemia.

Table 5.

Univariable and Multivariable Regression Analysis of Variables Associated with Severe Hypoglycemia Events in the Past Calendar Year in the 305 Participants

	Univariable analysis ^a		Multivariable analysis ^b
Characteristic	RR	P value	RR	P value
General
Age	1.03	<0.0001^c	1.02	0.04^c
Triglycerides	1.22	0.041^c	1.33	0.09
Problem Areas In Diabetes score	1.03	<0.0001^c	1.01	0.47
Geriatric Depression Scale score	1.20	<0.0001^c	1.07	0.05
Macrovascular complications (compared with none)	2.37	<0.0001^c	1.80	0.0004^c
Habits
Currently physically active (compared with no)	0.64	0.0004^c	1.10	0.60
Self-assessment
Daily frequency of capillary glucose tests	0.91	0.002^c	0.95	0.22
CGM use^d		0.001^c		0.001^c
Nonhabitual	0.69		1.82
Habitual	0.29		0.30
Hypoglycemia
Number of minor events in past calendar year (per month)	1.01	0.06	1.03	0.0001^c
HFS total	1.03	<0.0001^c	1.03	<0.0001^c
Insulin therapy
Use of an insulin pump	0.41	0.0001^c	0.50	<0.0001^c

The self-reported number of severe hypoglycemia events in the past year was analyzed by Poisson regression modeling. The risk ratio (RR) and P value are given for each variable.

All variables from Tables 1 –3 were analyzed in univariable Poisson regression, but only the significant results are showed in Table 5, including the number of minor hypoglycemia events in the past year.

All variables from univariable linear regression with P < 0.10 were included in the multivariable analysis.

Statistically significant difference, P < 0.05.

Nonhabitual continuous glucose monitoring (CGM) use was defined as <1 week per month, and habitual CGM use was defined as ≥1 weeks per month. The reference group consisted of participants who never adopted CGM use.

HFS, Hypoglycemia Fear Questionnaire score.

Discussion

Among questionnaire respondents living with T1DM for 50 years or more, although CGM use was uncommon, we found a proportion of insulin pump therapy use that exceeded expectations. Specifically, insulin pump use was reported in 44% of participants, whereas only 8% reported habitual CGM use (and a further 5% reported nonhabitual use). Although the pump subgroup reported both modifiable and nonmodifiable characteristics that implied better diabetes self-management skills—such as higher level of education, more frequent capillary glucose monitoring, more frequent use of a correction dose for hyperglycemia, and more frequent use of carbohydrate counting—pump use itself or the use of CGM was not independently associated with lower levels of HbA1c. Rather, implementation of a higher frequency of daily capillary glucose monitoring as a basic self-management skill was the key factor—independent of pump use and CGM—associated with lower levels of HbA1c. However, both of the technologies—insulin pump use and habitual CGM use—were each independently associated with lower risk of severe hypoglycemia reported in the year preceding participation in the study. These observational findings imply that two complementary factors—basic self-management skills and the use of technology—each play a key role in glycemic control and may be similarly applicable to older adults with long-standing T1DM as in the younger population with shorter diabetes duration in which glycemic benefit of these interventions has been demonstrated.

The proportion of insulin pump users shown in this Canadian study is higher than expected according to recent publications even when compared with other Western countries.³² In the United States, the country generally recognized for having the highest prevalence of use, insulin pump penetration has been estimated at 20–30% of patients with T1DM.³³ In Canada, the Alberta Institute of Health Economics reported that between 11,000 and 18,000 Canadians were using an insulin pump in 2009, representing less than 10% of the T1DM population at that time.⁵ These estimates arise from studies that generally evaluate younger patients³⁴ and include jurisdictions (such as the provinces of Ontario and Alberta in Canada, in which programs were initiated in 2008 and 2013, respectively) that have subsidized public health programs for the financial coverage of insulin pump therapy among adults with T1DM.³⁵ In most of the other provinces, insulin pump coverage is offered only to children and adolescents under the age of 18 years.

We hypothesize that the substantially higher proportion of insulin pump users observed in our cohort than these published comparators may arise from the increase in insulin pump use since the implementation of the public health programs. However, the higher proportion may also arise at least in part from two forms of selection bias: self-selection bias (commonly referred to as volunteer bias) and incidence–prevalence bias (commonly referred to as survival bias). The first arises from the potential for a participant's adoption of new technologies in the treatment of diabetes to be associated with his or her likelihood to participate in the study. The second arises when those with milder forms of diabetes associated with less morbidity are simultaneously more likely to adopt pump therapy and more likely to be sampled for study participation because those with more severe forms of disease are not available for sampling owing to death or impairment.³⁶

Although we could not directly estimate the impact of selection bias on the estimate of insulin pump prevalence, we could estimate a lower bound derived from the pump subgroup observed in our cohort divided by an estimated denominator of all Canadians with 50 or more years of T1DM estimated to be alive at the time of study. This estimate was derived from (1) knowledge of estimated diabetes prevalence in 1962, (2) anticipated general survival based on the 1962 Statistics Canada census,²⁹ and (3) a contemporaneous, age-specific survival estimates for T1DM.³⁰ From these data, with an estimate of 50% survival, we anticipate the existence of 1,441 Canadians living with T1DM for 50 years or more. As such, the most conservative estimate of insulin pump prevalence, observed in 133 members of the Longevity cohort, is 9%. This figure approaches the estimate for the entire population of Canadians with T1DM,⁵ implying that insulin pump use among those with 50 years or more of diabetes is at least as comparable in magnitude to the T1DM reference population of all ages. Although estimates of CGM use in older adults with long-standing diabetes or among Canadians are not known, we report what appears to be a very low adoption in this population in comparison with the 18% use observed in the American T1D Exchange Clinic Registry among adults 50 years of age or older with a mean diabetes duration of 33 years.³⁷

Although self-selection and incidence–prevalence bias may exaggerate the assessment of prevalent insulin pump use, for the purpose of the analytical comparison of factors associated with insulin pump use and glycemic control, we do not anticipate that these or other sources of selection bias are likely to affect internal validity. However, caution must be taken in interpreting the finding that the lower HbA1c level observed in the pump subgroup (–0.2%) was not statistically significant, implying a lack of clinical efficacy from insulin pump therapy. Unlike randomized clinical trials, in which key inclusion criteria included an HbA1c level above target before randomization of subjects to the intervention,^6

–9 the current nonexperimental study compared pump use in a cohort with near-target glycemic control. Furthermore, clinical characteristics simultaneously associated with insulin pump use and with glycemic control (potential confounding variables) may be unknown or unmeasured. Despite this substantial bias toward a null result, the magnitude of HbA1c difference between the two treatment groups approached what had been observed in previous clinical trials of younger patients with T1DM with higher baseline HbA1c level (–0.25% observed in meta-analysis).⁹

To our knowledge, only three studies have been conducted in patients with long-standing T1DM that aimed to determine the impact of insulin pump therapy on glycemic control. The first, a cross-sectional analysis, could not demonstrate a difference in HbA1c level between an elderly and a younger subgroup of participants treated with insulin pump therapy.³⁸ The second was a retrospective case series of five elderly patients who transitioned to insulin pump therapy and experienced a mean 2.0% decline in HbA1c level after 1 year (HbA1c, 9.6–7.6% at 1 year; P < 0.003).³⁹ The third was a prospective study of 34 subjects 50 years of age or more who were transitioned to insulin pump therapy and experienced a substantial 1-year decline in HbA1c level from 7.64% to 7.01% (P < 0.01).⁴⁰ Both longitudinal studies also showed a significant decrease in severe hypoglycemic episodes. Although the findings from these nonrandomized prospective studies suggest that insulin pump therapy benefits patients regardless of age and diabetes duration, we could not confirm from our cross-sectional analysis that pump use itself is independently associated with lower levels of HbA1c.

In keeping with previous reports, regardless of the use of insulin pump therapy, higher frequency of daily glucose monitoring was the only modifiable factor related to diabetes care that was independently associated with lower levels of HbA1c.^41,42 Specifically, each additional daily glucose test was associated with a 0.12% lower HbA1c. This finding implies a key lesson in the long-term enterprise of managing diabetes successfully: while in this study insulin pump therapy and CGM appear to have beneficial impact on reducing severe hypoglycemia events in those with long-standing T1DM, better adherence to simple self-assessment behavior itself appeared to outperform the use of advanced technology in the achievement of lower levels of HbA1c. We can conclude that basic self-management skill and the use of technology each play complementary roles in the two aspects of improving glycemic control: lowering HbA1c levels and preventing severe hypoglycemia.

While we report the first study investigating the prevalence and factors associated with insulin pump therapy in a large cohort of patients with long-standing T1DM, we recognize sources of selection bias for insulin pump prevalence (self-selection and incidence–prevalence bias) and therefore have reported a conservative estimate of prevalence as a sensitivity analysis. Although the analytical comparisons of factors associated with insulin pump and CGM use with measures of glycemic control were not subject to these selection biases, we recognize the presence of recall bias as a feature of the questionnaire study. We attempted to minimize this bias through the focus on current health status and treatment, such as the operationalization of events occurring in the year prior to study. To condition on potential confounding variables we performed adjusted analysis, and we also conducted a sensitivity analysis in which exclusion of participants with fewer than three injections per day did not influence results. Finally, we emphasize that this report presents cross-sectional baseline data, and as such it does not represent a prospective study, and we did not document clinical reasons for insulin pump initiation or CGM use. Although we could not demonstrate a cross-sectional association of insulin pump therapy and quality of life, longitudinal studies have shown improvement of quality of life indicators following pump initiation.⁴³

Although insulin pump use appeared frequent among patients with long-standing T1DM, its beneficial role, along with CGM, appears to be centered on the prevention of severe hypoglycemia. For lower HbA1c levels, the central independent modifiable factor identified was higher frequency of daily glucose monitoring. Although insulin pump and CGM technologies may be similarly beneficial for hypoglycemia prevention, these findings imply that they play a complementary role together with basic self-management skills. Consequently, strategies to facilitate glucose monitoring and pattern recognition must remain a key focus in the design and implementation of interventions for improving glycemic control regardless of the method of insulin delivery in older adults with long-standing T1DM.^44,45

Footnotes

Acknowledgments

We are grateful to the support of JDRF Canada, its Canadian Clinical Trial Network, and the Canadian Diabetes Association for their nonfinancial support with study advertisement.

Author Disclosure Statement

B.A.P. has received speaker honoraria from Medtronic Inc., Johnson and Johnson, Roche, Glaxo SmithKline Canada, Novo Nordisk, and Sanofi, has received research grant support from Medtronic and Boehringer Ingelheim, and serves as a consultant for Neurometrix. D.Z.I.C. has received speaker honoraria from Johnson and Johnson, Boehringer-Ingelheim, Merck, and AstraZeneca and has received research grant support from Merck, Boehringer-Ingelheim, and AstraZeneca. G.B., E.M.H., L.E.L., A.W., J.-W.B., D.E., H.A.K., M.H.B., N.P., and V.B. have no disclosures.

G.B. and B.A.P. researched data and wrote the manuscript as primary authors. E.M.H., L.E.L., J.-W.B., and D.E. researched data, contributed to discussion, and reviewed/edited the manuscript. A.W., H.A.K., M.H.B., N.P., V.B., and D.Z.I.C. contributed to discussion and reviewed/edited the manuscript. All authors have approved the final version of this manuscript.

References

The Diabetes Control and Complications Trial Research Group: The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Engl J Med, 1993; 329:977–986.

Gregg

, Li

, Wang

, et al.: Changes in diabetes-related complications in the United States, 1990–2010. N Engl J Med, 2014; 370:1514–1523.

Klein

, Klein

, Moss

, et al.: The Wisconsin Epidemiologic Study of Diabetic Retinopathy. III. Prevalence and risk of diabetic retinopathy when age at diagnosis is 30 or more years. Arch Ophthalmol, 1984; 102:527–532.

Sustained effect of intensive treatment of type 1 diabetes mellitus on development and progression of diabetic nephropathy: the Epidemiology of Diabetes Interventions and Complications (EDIC) study. JAMA, 2003; 290:2159–2167.

Institute of Health Economics: Insulin Pump Therapy for Type 1 Diabetes. Alberta STE Report. Edmonton, AB, Canada: Institute of Health Economics, 2012:26–28.

Weissberg-Benchell

, Antisdel-Lomaglio

, Seshadri

: Insulin pump therapy: a meta-analysis. Diabetes Care, 2003; 26:1079–1087.

Jakisch

, Wagner

, Heidtmann

, et al.: Comparison of continuous subcutaneous insulin infusion (CSII) and multiple daily injections (MDI) in paediatric Type 1 diabetes: a multicentre matched-pair cohort analysis over 3 years. Diabet Med, 2008; 25:80–85.

Hoogma

, Hammond

, Gomis

, et al.: Comparison of the effects of continuous subcutaneous insulin infusion (CSII) and NPH-based multiple daily insulin injections (MDI) on glycaemic control and quality of life: results of the 5-Nations Trial. Diabet Med, 2006; 23:141–147.

Misso

, Egberts

, Page

, et al.: Continuous subcutaneous insulin infusion (CSII) versus multiple insulin injections for type 1 diabetes mellitus. Cochrane Database Syst Rev, 2010;(1):CD005103.

10.

Didangelos

, Iliadis

: Insulin pump therapy in adults. Diabetes Res Clin Pract, 2011; 93(Suppl 1):S109–S113.

11.

Pickup

, Sutton

: Severe hypoglycaemia and glycaemic control in Type 1 diabetes: meta-analysis of multiple daily insulin injections compared with continuous subcutaneous insulin infusion. Diabet Med, 2008; 25:765–774.

12.

Walsh

, Roberts

: Pumping Insulin: Everything You Need to Succeed on an Insulin Pump, 5 ^th ed. La Jolla, CA: Torrey Pines Press, 2012.

13.

Schutt

, Fach

, Seufert

, et al.: Multiple complications and frequent severe hypoglycaemia in ‘elderly’ and ‘old’ patients with Type 1 diabetes. Diabet Med, 2012; 29:e176–e179.

14.

Boland

, Grey

, Oesterle

, et al.: Continuous subcutaneous insulin infusion. A new way to lower risk of severe hypoglycemia, improve metabolic control, and enhance coping in adolescents with type 1 diabetes. Diabetes Care, 1999; 22:1779–1784.

15.

Brink

, Stewart

: Insulin pump treatment in insulin-dependent diabetes mellitus. Children, adolescents, and young adults. JAMA, 1986; 255:617–621.

16.

Weisman

, Lovblom

, Rovinski

, et al.: Neuropathy prevalence compared with other complications in longstanding T1DM: preliminary analysis of the Canadian Study of Longevity in Diabetes cohort [abstract 1580-P]. Diabetes, 2015; 64(Suppl 1):A411.

17.

Bai

, Boulet

, Halpern

, et al.: Cardiovascular disease guideline adherence and self-reported statin use in longstanding type 1 diabetes: results from the Canadian Study of Longevity in Diabetes cohort. Cardiovasc Diabetol, 2016; 15:14.

18.

Keenan

, Costacou

, Sun

, et al.: Clinical factors associated with resistance to microvascular complications in diabetic patients of extreme disease duration: the 50-year medalist study. Diabetes Care, 2007; 30:1995–1997.

19.

Bain

, Gill

, Dyer

, et al.: Characteristics of Type 1 diabetes of over 50 years duration (the Golden Years Cohort). Diabet Med, 2003; 20:808–811.

20.

Levey

, Stevens

, Schmid

, et al.: A new equation to estimate glomerular filtration rate. Ann Intern Med, 2009; 150:604–612.

21.

Montorio

, Izal

: The Geriatric Depression Scale: a review of its development and utility. Int Psychogeriatr, 1996; 8:103–112.

22.

Wancata

, Alexandrowicz

, Marquart

, et al.: The criterion validity of the Geriatric Depression Scale: a systematic review. Acta Psychiatr Scand, 2006; 114:398–410.

23.

Welch

, Jacobson

, Polonsky

: The Problem Areas in Diabetes Scale. An evaluation of its clinical utility. Diabetes Care, 1997; 20:760–766.

24.

Moghtaderi

, Bakhshipour

, Rashidi

: Validation of Michigan neuropathy screening instrument for diabetic peripheral neuropathy. Clin Neurol Neurosurg, 2006; 108:477–481.

25.

Feldman

, Stevens

, Thomas

, et al.: A practical two-step quantitative clinical and electrophysiological assessment for the diagnosis and staging of diabetic neuropathy. Diabetes Care, 1994; 17:1281–1289.

26.

Cheng

: Canadian Diabetes Association 2013 clinical practice guidelines for the prevention and management of diabetes in Canada. Introduction. Can J Diabetes, 2013; 37(Suppl 1):S1–S3.

27.

Sun

, Keenan

, Cavallerano

, et al.: Protection from retinopathy and other complications in patients with type 1 diabetes of extreme duration: the Joslin 50-year Medalist Study. Diabetes Care, 2011; 34:968–974.

28.

Gonder-Frederick

, Schmidt

, Vajda

, et al.: Psychometric properties of the Hypoglycemia Fear Survey-II for adults with type 1 diabetes. Diabetes Care, 2011; 34:801–806.

29.

Harris

, Worrall

, Macaulay

, et al.: Diabetes management in Canada: baseline results of the Group Practice Diabetes Management Study. Can J Diabetes, 2006; 30:131–137.

30.

Krolewski

, Warram

: Natural history of diabetes mellitus. In: Becker KL, ed. Principles and Practice of Endocrinology and Metabolism. Philadelphia: J.B. Lippincott Co., 1995:1216–1220.

31.

Pambianco

, Costacou

, Ellis

, et al.: The 30-year natural history of type 1 diabetes complications: the Pittsburgh Epidemiology of Diabetes Complications Study experience. Diabetes, 2006; 55:1463–1469.

32.

White

, Goenka

, Furlong

, et al.: The U.K. service level audit of insulin pump therapy in adults. Diabet Med, 2014; 31:412–418.

33.

HSBC Global Research: Diabetes: Proprietary Survey on Insulin Pumps and Continuous Blood Glucose Monitoring.. Healthcare US Equipment & Supplies, 2005. www.research.hsbc.com/midas/Res/RDV?p=pdf&key=y8f6w0euwd&name=121747.PDf (accessed February 1, 2015 ).

34.

Pickup

, Hammond

: NICE guidance on continuous subcutaneous insulin infusion 2008: review of the technology appraisal guidance. Diabet Med, 2009; 26:1–4.

35.

Canadian Diabetes Association: Access to Medications, Devices and Supplies & Your Rights. 2014. www.diabetes.ca/diabetes-and-you/know-your-rights/access-to-medications-devices-supplies-your-rights (accessed February 1, 2015 ).

36.

Rothman

, Greenland

, Lash

: Modern Epidemiology. Philadelphia: Wolters Kluwer/Lippincott Williams & Wilkins, 2012.

37.

Miller

, Foster

, Beck

, et al.: Current state of type 1 diabetes treatment in the U.S.: updated data from the T1D Exchange clinic registry. Diabetes Care, 2015; 38:971–978.

38.

Matejko

, Cyganek

, Katra

, et al.: Insulin pump therapy is equally effective and safe in elderly and young type 1 diabetes patients. Rev Diabet Stud, 2011; 8:254–258.

39.

Rizvi

, Petry

, Arnold

, et al.: Beneficial effects of continuous subcutaneous insulin infusion in older patients with long-standing type 1 diabetes. Endocr Pract, 2001; 7:364–369.

40.

Siegel-Czarkowski

, Herold

, Goland

: Continuous subcutaneous insulin infusion in older patients with type 1 diabetes. Diabetes Care, 2004; 27:3022–3023.

41.

Shalitin

, Gil

, Nimri

, et al.: Predictors of glycaemic control in patients with Type 1 diabetes commencing continuous subcutaneous insulin infusion therapy. Diabet Med, 2010; 27:339–347.

42.

The DCCT Research Group: Epidemiology of severe hypoglycemia in the Diabetes Control and Complications Trial. Am J Med, 1991; 90:450–459.

43.

Linkeschova

, Raoul

, Bott

, et al.: Less severe hypoglycaemia, better metabolic control, and improved quality of life in Type 1 diabetes mellitus with continuous subcutaneous insulin infusion (CSII) therapy; an observational study of 100 consecutive patients followed for a mean of 2 years. Diabet Med, 2002; 19:746–751.

44.

New

, Ajjan

, Pfeiffer

, et al.: Continuous glucose monitoring in people with diabetes: the randomized controlled Glucose Level Awareness in Diabetes Study (GLADIS). Diabet Med, 2015; 32:609–617.

45.

Thomas

, Kane

, Bakst

, et al.: A glucose meter accuracy and precision comparison: the FreeStyle Flash Versus the Accu-Chek Advantage, Accu-Chek Compact Plus, Ascensia Contour, and the BD Logic. Diabetes Technol Ther, 2008; 10:102–110.