To describe rescue insulin use and associated factors in the Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE).
GRADE participants (type 2 diabetes duration <10 years, baseline A1C 6.8%–8.5% on metformin monotherapy, N = 5,047) were randomly assigned to insulin glargine U-100, glimepiride, liraglutide, or sitagliptin and followed quarterly for a mean of 5 years. Rescue insulin (glargine or aspart) was to be started within 6 weeks of A1C >7.5%, confirmed. Reasons for delaying rescue insulin were reported by staff-completed survey.
Nearly one-half of GRADE participants (N = 2,387 [47.3%]) met the threshold for rescue insulin. Among participants assigned to glimepiride, liraglutide, or sitagliptin, rescue glargine was added by 69% (39% within 6 weeks). Rescue aspart was added by 44% of glargine-assigned participants (19% within 6 weeks) and by 30% of non-glargine-assigned participants (14% within 6 weeks). Higher A1C values were associated with adding rescue insulin. Intention to change health behaviors (diet/lifestyle, adherence to current treatment) and not wanting to take insulin were among the most common reasons reported for not adding rescue insulin within 6 weeks.
Proportionately, rescue glargine, when required, was more often used than rescue aspart, and higher A1C values were associated with greater rescue insulin use. Wanting to use noninsulin strategies to improve glycemia was commonly reported, although multiple factors likely contributed to not using rescue insulin. These findings highlight the persistent challenge of intensifying type 2 diabetes treatment with insulin, even in a clinical trial.
Introduction
Delaying insulin treatment for type 2 diabetes is common and is often cited as a specific example of “therapeutic inertia,” that is, not changing treatment despite the availability of guidelines to do so, awareness of the guidelines by the provider, applicability of the guidelines to the individual patient, and availability of resources to apply the guidelines (1–4). Patient-related factors contributing to therapeutic inertia include fear of hypoglycemia, fear of injections, lack of resources, and associating the need for insulin with personal failure; provider-related factors include fear of side effects, treatment complexity, ignorance of treatment guidelines, and the belief that insulin should be used only as a treatment of last resort, while system-related factors include access to care and resource limitations and restrictions (1–3,5–9).
The Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE) protocol provided criteria and instructions for adding “rescue” insulin to metformin and the randomly assigned treatment. In this descriptive analysis, we examine associations of various baseline and clinical factors with rescue insulin use and describe the effects of rescue insulin on glycemia, weight, and the incidence of hypoglycemia. Reasons for not adding rescue insulin as assessed with a retrospectively developed, staff-completed survey are also presented.
Research Design and Methods
GRADE was a multicenter, parallel-arm comparative effectiveness clinical trial comparing four major classes of glucose-lowering medications in people with type 2 diabetes diagnosed for <10 years with baseline A1C of 6.8%–8.5% (51–69 mmol/mol) and on maximally tolerated doses of metformin (N = 5,047) (10). The full protocol is available from https://grade.bsc.gwu.edu. The study was conducted at 36 clinical centers across the U.S. (11). The randomly assigned study medication, insulin glargine U-100 (hereafter, glargine), the sulfonylurea glimepiride, the glucagon-like peptide 1 (GLP-1) receptor agonist liraglutide, or the dipeptidyl peptidase 4 inhibitor sitagliptin, was added to the participant’s maximally tolerated metformin dose, up to 2,000 mg daily (mean ± SD dose 1,994 ± 205 mg at baseline) (11). Written informed consent, indicating willingness to accept random treatment assignment and to add glargine, and/or prandial insulin (insulin aspart U-100 [aspart]) in response to persistent A1C >7.5% (>58 mmol/mol), was obtained from all participants. Enrolled participants were evaluated at quarterly visits over a mean follow-up of 5 years (11).
The GRADE primary metabolic outcome was time to A1C ≥7.0% (≥53 mmol/mol), confirmed at the subsequent quarterly visit while participants were on maximally tolerated doses of metformin and the assigned medication. The secondary metabolic outcome (hereafter secondary outcome) was the time to A1C >7.5% (>58 mmol/mol), confirmed at the subsequent quarterly visit, and the tertiary metabolic outcome (hereafter tertiary outcome) was time to A1C >7.5% (>58 mmol/mol), confirmed at the subsequent visit for participants who had already met the secondary outcome (11). The GRADE protocol specified that rescue insulin be started after confirmation of the secondary and tertiary outcomes (operationalized as within 6 weeks of outcome confirmation). Rescue insulin following confirmation of the secondary outcome was a daily injection of glargine added to glimepiride, liraglutide, or sitagliptin (non-glargine-assigned groups) or a daily injection of aspart added to glargine (glargine-assigned group). If the tertiary outcome was subsequently confirmed, non-glargine-assigned participants were to discontinue the randomly assigned treatment, continue metformin plus glargine, and add a single daily injection of aspart. Glargine-assigned participants continued on metformin, glargine, and aspart. Insulin dose adjustments, including progression to multiple daily injections of aspart, were made according to protocol guidelines and as informed by participants’ self-monitored blood glucose results.
Rescue insulin use across clinical sites was monitored centrally by the GRADE Protocol Oversight Committee. The Non-Initiation of Insulin (NONINIT) survey was developed by the committee to understand the reasons for delaying rescue insulin beyond 6 weeks of outcome confirmation and to identify any clinical site variability in rescue insulin use. The NONINIT survey included a list of 52 possible reasons for not adding rescue insulin, addressing participant, provider, and system-related factors that contribute to therapeutic inertia (3,6). Clinical site staff selected and rank ordered the four most influential reasons based on discussions with the participant and relevant clinical factors. The survey was implemented in 2015, ∼2.5 years after study initiation. Surveys were completed retrospectively for those participants who had already met the secondary (N = 86) or tertiary (N = 24) outcome.
Statistical Methods
The analysis of rescue insulin use following the secondary outcome includes 2,387 GRADE participants (47% of the study cohort). Participants who added rescue insulin prior to confirmation of the secondary outcome (N = 21) or were missing outcome confirmation forms (N = 3) are excluded. The analysis of rescue insulin use following the tertiary outcome is limited to non-glargine-assigned participants who added glargine after confirmation of the secondary outcome (N = 803 [16% of the GRADE cohort]). Non-glargine-assigned participants who never added glargine after the secondary outcome (N = 292), and glargine-assigned participants (N = 497) for whom aspart should have already been added following the secondary outcome, are excluded. Participants who added rescue aspart prior to confirming the tertiary outcome (N = 4), or were missing outcome confirmation forms (N = 2), are also excluded (Supplementary Fig. 1).
Baseline and demographic characteristics of the GRADE study cohort have previously been described (11,12). Baseline characteristics selected for this analysis include sex (male, female), race (White, Black, other), ethnicity (Hispanic, non-Hispanic), and treatment assignment (glargine, glimepiride, liraglutide, sitagliptin). Additional characteristics include age (years) and diabetes duration (years) at the time of the relevant outcome (secondary, tertiary), time (years) from baseline to the outcome, A1C (%) and body weight (kilograms) at outcome confirmation, and incidence of self-reported hypoglycemia (any vs. none) occurring between baseline and the outcome. Hypoglycemia did not require confirmation by glucose testing. Time to adding rescue insulin (“time-to-add”) was defined categorically as follows: within 6 weeks (≤6 weeks) of outcome confirmation, more than 6 weeks but prior to study end (>6 weeks), or never.
Participant characteristics were summarized with counts and percentages and means ± SD. A Classification And Regression Tree (CART) model was built using the following as predictor variables: baseline variables (sex, race, Hispanic ethnicity, treatment assignment, and study site), variables at time of the outcome (age, diabetes duration, weight, and A1C at the outcome confirmation visit), and incident hypoglycemia (any vs. none based on self-report) preceding the relevant outcome. The CART model algorithm identified which factors were predictive of adding insulin at ≤6 weeks and calculated the probability of adding insulin at ≤6 weeks for different categories of the identified predictor variables. Cross validation was used to prune the CART model to prevent overfitting the model to the data, and the model minimizing the cross-validation error was selected (13).
Differences in change in A1C and body weight (F test from a linear regression model), and incidence of hypoglycemia (χ2 test from a logistic regression model) among the three time-to-add categories, were examined, with time-to-add as the only independent variable in the model. If the global test of differences across the three time-to-add categories was significant, then tests of pairwise differences were conducted, and a closed testing procedure was used to control type 1 error due to multiple comparisons (14). The global and pairwise tests of differences were conducted at 2 years after the outcome to maximize the analytic sample size, as the number of participants with >2 years of follow-up after outcome confirmation decreases considerably.
The NONINIT survey results are summarized with counts and stratified by rescue insulin type (glargine [non-glargine-assigned participants after the secondary outcome] or aspart [glargine-assigned participants after the secondary outcome, non-glargine-assigned participants after the tertiary outcome]). Only the most common responses (those accounting for at least 95% of the NONINIT reasons and listed as the primary reason for ≥10 participants) are shown.
Data and Resource Availability
This article is based on data collected at baseline and follow-up and outcome assessments from the 5,047 participants enrolled in GRADE, the database for which will be available in the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Central Repository in 2024.
Results
Rescue Insulin Following the Secondary Outcome
Among the 2,387 participants reaching the secondary outcome, rescue insulin was added at ≤6 weeks by 35% (N = 834, mean ± SD 16 ± 11 days) and >6 weeks by 29% (N = 698, 301 ± 299 days) and never added by 36% (N = 855) (Table 1). Cumulative incidence curves of time-to-add up to 1 year after reaching the secondary outcome are shown (Fig. 1).
. | Total secondary outcomes . | Time-to-add (glargine/aspart) . | |||
---|---|---|---|---|---|
≤6 weeks . | >6 weeks . | Never . | P . | ||
N total | 2,387 | 834 | 698 | 855 | |
Sex | 0.112 | ||||
Male | 1,512 | 536 (35.4) | 420 (27.8) | 556 (36.8) | |
Female | 875 | 298 (34.1) | 278 (31.8) | 299 (34.2) | |
Race | <0.001 | ||||
White | 1,540 | 595 (38.6) | 423 (27.5) | 522 (33.9) | |
Black | 460 | 118 (25.7) | 142 (30.9) | 200 (43.5) | |
Other | 387 | 121 (31.3) | 133 (34.4) | 133 (34.4) | |
Ethnicity | 0.416 | ||||
Non-Hispanic | 1,866 | 658 (35.3) | 533 (28.6) | 675 (36.2) | |
Hispanic | 507 | 173 (34.1) | 160 (31.6) | 174 (34.3) | |
Treatment group | <0.001 | ||||
Glargine | 497 | 93 (18.7) | 128 (25.8) | 276 (55.5) | |
Glimepiride | 625 | 213 (34.1) | 186 (30.0) | 226 (36.2) | |
Liraglutide | 572 | 194 (33.9) | 188 (32.9) | 190 (33.2) | |
Sitagliptin | 693 | 334 (48.2) | 196 (28.3) | 163 (23.5) | |
Age at outcome (years) | 57.1 ± 10.3 | 56.5 ± 10.3 | 55.5 ± 10.2 | 58.9 ± 10.2 | <0.001 |
Diabetes duration at outcome (years) | 6.2 ± 3.0 | 6.0 ± 2.9 | 5.9 ± 2.9 | 6.8 ± 3.0 | <0.001 |
Time to secondary outcome (years) | 2.2 ± 1.5 | 2.0 ± 1.4 | 1.9 ± 1.2 | 2.8 ± 1.5 | <0.001 |
Confirmation A1C (%) at secondary outcome | 8. 4 ±1.0 | 8.6 ± 1.1 | 8.4 ± 0.9 | 8.3 ± 0.9 | <0.001 |
Confirmation A1C (mmol/mol) at secondary outcome | 68 ± 10.9 | 70 ± 12.0 | 68 ± 9.8 | 67 ± 9.8 | <0.001 |
Weight at secondary outcome (kg) | 101.2 ± 22.8 | 103.7 ± 24.2 | 100.9 ± 22.6 | 98.8 ± 21.2 | <0.001 |
Incidence of hypoglycemia† | 1,110 | 375 (33.8) | 311 (28.0) | 424 (38.2) | 0.077 |
Time to adding insulin (days) | 145.8 ± 246.9 | 15.9 ± 11.2 | 300.9 ± 299.2 | NA | <0.001 |
. | Total secondary outcomes . | Time-to-add (glargine/aspart) . | |||
---|---|---|---|---|---|
≤6 weeks . | >6 weeks . | Never . | P . | ||
N total | 2,387 | 834 | 698 | 855 | |
Sex | 0.112 | ||||
Male | 1,512 | 536 (35.4) | 420 (27.8) | 556 (36.8) | |
Female | 875 | 298 (34.1) | 278 (31.8) | 299 (34.2) | |
Race | <0.001 | ||||
White | 1,540 | 595 (38.6) | 423 (27.5) | 522 (33.9) | |
Black | 460 | 118 (25.7) | 142 (30.9) | 200 (43.5) | |
Other | 387 | 121 (31.3) | 133 (34.4) | 133 (34.4) | |
Ethnicity | 0.416 | ||||
Non-Hispanic | 1,866 | 658 (35.3) | 533 (28.6) | 675 (36.2) | |
Hispanic | 507 | 173 (34.1) | 160 (31.6) | 174 (34.3) | |
Treatment group | <0.001 | ||||
Glargine | 497 | 93 (18.7) | 128 (25.8) | 276 (55.5) | |
Glimepiride | 625 | 213 (34.1) | 186 (30.0) | 226 (36.2) | |
Liraglutide | 572 | 194 (33.9) | 188 (32.9) | 190 (33.2) | |
Sitagliptin | 693 | 334 (48.2) | 196 (28.3) | 163 (23.5) | |
Age at outcome (years) | 57.1 ± 10.3 | 56.5 ± 10.3 | 55.5 ± 10.2 | 58.9 ± 10.2 | <0.001 |
Diabetes duration at outcome (years) | 6.2 ± 3.0 | 6.0 ± 2.9 | 5.9 ± 2.9 | 6.8 ± 3.0 | <0.001 |
Time to secondary outcome (years) | 2.2 ± 1.5 | 2.0 ± 1.4 | 1.9 ± 1.2 | 2.8 ± 1.5 | <0.001 |
Confirmation A1C (%) at secondary outcome | 8. 4 ±1.0 | 8.6 ± 1.1 | 8.4 ± 0.9 | 8.3 ± 0.9 | <0.001 |
Confirmation A1C (mmol/mol) at secondary outcome | 68 ± 10.9 | 70 ± 12.0 | 68 ± 9.8 | 67 ± 9.8 | <0.001 |
Weight at secondary outcome (kg) | 101.2 ± 22.8 | 103.7 ± 24.2 | 100.9 ± 22.6 | 98.8 ± 21.2 | <0.001 |
Incidence of hypoglycemia† | 1,110 | 375 (33.8) | 311 (28.0) | 424 (38.2) | 0.077 |
Time to adding insulin (days) | 145.8 ± 246.9 | 15.9 ± 11.2 | 300.9 ± 299.2 | NA | <0.001 |
Data are means ± SD for continuous variables and numbers (row %) for categorical variables. Note that the protocol specified that, after the secondary outcome (A1C >7.5% [>58 mmol/mol], confirmed at the subsequent quarterly visit), glargine insulin should be added for participants randomized to glimepiride, liraglutide, or sitagliptin, and aspart insulin should be added for participants randomized to glargine. NA, not applicable.
†Incidence of hypoglycemia is defined as at least one self-reported hypoglycemia episode between baseline and the time of the outcome confirmation.
Participant characteristics stratified by the three time-to-add categories are shown in Table 1. A greater proportion of White participants (39%) compared with Black participants (26%) and others (31%) added insulin at ≤6 weeks, while a greater proportion of Black participants (44%) compared with White (34%) or others (34%) never added rescue insulin (P < 0.001). Rescue insulin use was highest among participants assigned to sitagliptin (77% overall, 48% at ≤6 weeks) and lowest among those assigned to glargine (45% overall, 19% at ≤6 weeks) (P < 0.001) (Table 1). Participants were older (mean ± SD age 58.9 ± 10.2 years) and diabetes duration was longest (6.8 ± 3.0 years) and time to reach the secondary outcome longest (2.8 ± 1.5 years) among those never using rescue insulin. The outcome A1C was highest (8.6 ± 1.1% [70 ± 12 mmol/mol]) among participants who added rescue insulin at ≤6 weeks (P < 0.001) (Table 1). From the CART model, the probability of adding rescue insulin at ≤6 weeks was highest (61%) among those assigned to sitagliptin with an outcome A1C >8.2% (66 mmol/mol) (Supplementary Fig. 2). Clinical site did not emerge as a significant factor.
Longitudinal graphs showing changes in A1C, body weight, and incident hypoglycemia by treatment group and time-to-add categories are shown in Supplementary Fig. 3. Two years after the secondary outcome, A1C improved in all three time-to-add categories (average A1C percentage point decline of 0.97 [10.6 mmol/mol], 0.41 [4.5 mmol/mol], and 0.46 [5.0 mmol/mol] for the ≤6 weeks, >6 weeks, and never groups, respectively). Weight increased among those adding insulin at ≤6 weeks (1.10 kg mean weight gain) and decreased in the >6 weeks (−0.2 kg) and never (−2.23 kg) groups. Incident hypoglycemia was reported by 25% (≤6 weeks), 16% (>6 weeks) and 14% (never) (data not shown). Differences in the changes in A1C, weight, and incidence of hypoglycemia among the three time-to-add groups at 2 years after secondary outcome confirmation were significant for all comparisons (P < 0.01) (Table 2).
Insulin initiation categories . | Difference or OR (SE) . | P . |
---|---|---|
Secondary outcome | ||
A1C change (%)a,c | <0.01 | |
≤6 vs. >6 weeksb | −0.57 (0.09) | <0.01 |
≤6 weeks vs. neverb | −0.51 (0.10) | <0.01 |
>6 weeks vs. neverb | 0.06 (0.11) | 0.58 |
Weight change (kg)a | <0.01 | |
≤6 vs. >6 weeksb | 1.30 (0.42) | <0.01 |
≤6 weeks vs. neverb | 3.33 (0.49) | <0.01 |
>6 weeks vs. neverb | 2.04 (0.51) | <0.01 |
Incidence of hypoglycemiaa,d | <0.01 | |
≤6 vs. >6 weeksb | 1.82 (0.29) | <0.01 |
≤6 weeks vs. neverb | 2.15 (0.42) | <0.01 |
>6 weeks vs. neverb | 1.19 (0.25) | 0.41 |
Tertiary outcome | ||
A1C change (%)a,c | <0.01 | |
≤6 vs. >6 weeksb | −0.68 (0.27) | 0.01 |
≤6 weeks vs. neverb | 0.31 (0.24) | 0.21 |
>6 weeks vs. neverb | 0.99 (0.21) | <0.01 |
Weight change (kg)a | 0.01 | |
≤6 vs. >6 weeksb | −2.48 (1.23) | 0.05 |
≤6 weeks vs. neverb | 0.29 (1.09) | 0.79 |
>6 weeks vs. neverb | 2.77 (0.92) | 0.01 |
Incidence of hypoglycemiaa,d | 0.01 | |
≤6 vs. >6 weeksb | 1.78 (0.69) | 0.14 |
≤6 weeks vs. neverb | 2.99 (1.05) | 0.01 |
>6 weeks vs. neverb | 1.68 (0.57) | 0.12 |
Insulin initiation categories . | Difference or OR (SE) . | P . |
---|---|---|
Secondary outcome | ||
A1C change (%)a,c | <0.01 | |
≤6 vs. >6 weeksb | −0.57 (0.09) | <0.01 |
≤6 weeks vs. neverb | −0.51 (0.10) | <0.01 |
>6 weeks vs. neverb | 0.06 (0.11) | 0.58 |
Weight change (kg)a | <0.01 | |
≤6 vs. >6 weeksb | 1.30 (0.42) | <0.01 |
≤6 weeks vs. neverb | 3.33 (0.49) | <0.01 |
>6 weeks vs. neverb | 2.04 (0.51) | <0.01 |
Incidence of hypoglycemiaa,d | <0.01 | |
≤6 vs. >6 weeksb | 1.82 (0.29) | <0.01 |
≤6 weeks vs. neverb | 2.15 (0.42) | <0.01 |
>6 weeks vs. neverb | 1.19 (0.25) | 0.41 |
Tertiary outcome | ||
A1C change (%)a,c | <0.01 | |
≤6 vs. >6 weeksb | −0.68 (0.27) | 0.01 |
≤6 weeks vs. neverb | 0.31 (0.24) | 0.21 |
>6 weeks vs. neverb | 0.99 (0.21) | <0.01 |
Weight change (kg)a | 0.01 | |
≤6 vs. >6 weeksb | −2.48 (1.23) | 0.05 |
≤6 weeks vs. neverb | 0.29 (1.09) | 0.79 |
>6 weeks vs. neverb | 2.77 (0.92) | 0.01 |
Incidence of hypoglycemiaa,d | 0.01 | |
≤6 vs. >6 weeksb | 1.78 (0.69) | 0.14 |
≤6 weeks vs. neverb | 2.99 (1.05) | 0.01 |
>6 weeks vs. neverb | 1.68 (0.57) | 0.12 |
OR, odds ratio. Secondary outcome: A1C >7.5% (>58 mmol/mol), confirmed at the subsequent quarterly visit. Tertiary outcome: A1C >7.5% (>58 mmol/mol), confirmed at the subsequent visit for participants who had already met the secondary outcome.
aP value is from a global test of differences across the three categories (F test from a linear regression for A1C change and weight change, χ2 test from a logistic regression for incidence of hypoglycemia).
bDifference (A1C change, weight change) or OR (incidence of hypoglycemia) for comparison of the first category with the second category (e.g., difference for ≤6 weeks category minus >6 weeks category, odds for ≤6 weeks category divided by the odds for the >6 weeks category). P value is from a pairwise test of differences between the two specific categories (F test from a linear regression for A1C change and weight change, χ2 test from a logistic regression for incidence of hypoglycemia). A closed testing procedure was used to control type 1 error due to multiple comparisons among the three pairwise comparisons.
cA1C values in % units can be converted to mmol/mol with the NGSP A1C converter, available from https://ngsp.org/convert1.asp.
dIncidence of at least one episode of self-reported hypoglycemia in the past 30 days.
Nearly one-half of participants (N = 1,002 [45%]) who confirmed the secondary outcome had an A1C ≤7.5% (≤58 mmol/mol) at the last study visit. The last study visit A1C was <7.0% (<53 mmol/mol) for 21% (N = 491). The likelihood of having an A1C <7% (<53 mmol/mol) was greatest among participants who added rescue insulin at ≤6 weeks. The likelihood of an A1C >7.5% (>58 mmol/mol) at the last study visit was greatest among those who never added rescue insulin (P < 0.001) (Supplementary Table 1).
Rescue Insulin Following the Tertiary Outcome
Following tertiary outcome confirmation (N = 803), rescue insulin aspart was added in ≤6 weeks by 14% (N = 109), in >6 weeks by 16% (N = 128), and never added by 70% (N = 566) (Fig. 1 and Supplementary Table 2). Characteristics associated with the time-to-add categories were similar for the tertiary and secondary outcome, except for an association with ethnicity rather than race (Supplementary Table 2). CART modeling identified only time to reach the tertiary outcome and body weight at the tertiary outcome as predictive of adding rescue insulin at ≤6 weeks. Specifically, among non-glargine-assigned participants who reached the tertiary outcome within 1 year of baseline (N = 47), the probability of aspart being added in ≤6 weeks was 24% for participants weighing ≥87 kg at the tertiary outcome (N = 38) and 100% for those weighing <87 kg (N = 9). The probability of aspart being added in ≤6 weeks was 12% among participants reaching the tertiary outcome >1 year from baseline, regardless of body weight (N = 756 [data not shown]). Changes in A1C, body weight, and incident hypoglycemia after the tertiary outcome by treatment group and time-to-add category are shown in Supplementary Fig. 4. Waiting to add rescue aspart beyond 6 weeks was associated with worsened A1C relative to the ≤6 week group (P = 0.01) and the never group (P < 0.01). The A1C change in the ≤6 week compared with the never group was not significantly different. Weight was increased in the >6 week relative to the ≤6 week group (P = 0.05) and the never group (P = 0.01). The odds of incident hypoglycemia were greater in the ≤6 week group compared with never (P = 0.01) (Table 2). A1C at the last study visit was ≤7.5% (≤58 mmol/mol) in 32% (N = 238) of participants, including 179 who never added rescue aspart. The last study visit A1C was <7% (<53 mmol/mol) in 11% (N = 92) participants, of whom 60 never added rescue aspart (Supplementary Table 1).
Rescue Insulin Use by Insulin Type (Glargine, Aspart)
During GRADE, 1,890 (50%) of all non-glargine-assigned participants (N = 3,784) met the secondary outcome, and 1,311 (69%) initiated rescue glargine (Table 1). The tertiary outcome was reached by 1,300 (34%) of non-glargine-assigned participants, and rescue aspart was initiated by 458 (35%) (Supplementary Table 2). Among all glargine-assigned participants (N = 1,263), 497 (39%) met the secondary outcome, of whom rescue aspart was initiated by 221 (44%) (Table 1).
NONINIT Survey Results
The most frequently recorded primary reasons for not adding rescue insulin at ≤6 weeks are shown in Fig. 2. Noninsulin glucose improvement strategies (e.g., making dietary or lifestyle changes, improving adherence to current medication treatment) were among the most common reasons. “Does not want to take insulin” was recorded for 201 and 11 non-glargine-assigned participants after the secondary and tertiary outcomes, respectively.
Conclusions
Since its introduction a century ago, insulin has played a major role in the treatment of diabetes. In GRADE, the randomly assigned treatments, including glargine, were initiated by 99.8% of participants. Discontinuation of the assigned treatment was infrequent and lowest among those assigned to glargine (11). In contrast, glargine as rescue insulin was initiated by only 69% of expected participants and within 6 weeks by just 39%. Rescue aspart was implemented less frequently and was more often delayed. Many studies have reported long delays, or participants not intensifying treatment at rates similar to, or even in excess of, those observed in GRADE (4,11,15–18). Yet the proportion of GRADE participants in whom rescue insulin was delayed or never used was higher than expected in a well-resourced study, especially given the near universal acceptance of glargine at baseline.
Strategies that may have contributed to glargine acceptance at baseline included a two-phase consent process (run-in, clinical trial), patient interviews with different members of the clinical staff, and provision of educational materials addressing the benefits, risks, and side effects of each medication and addressing common misconceptions about insulin (19,20). Patients were required to demonstrate self-glucose monitoring and self-injection, the latter a proven strategy for moving people to accept insulin treatment (20). Efforts to improve rescue insulin use during GRADE included monthly site-level reporting of actual versus expected rates of rescue insulin use, outreach by study leadership to site investigators to encourage protocol adherence, and development and implementation of the NONINIT survey. Strategies to promote participant engagement in GRADE generally, and rescue insulin use particularly, were discussed at central meetings and during biweekly study team teleconferences. Clinical staff were encouraged to engage participants “early and often” in discussions of the possible need for insulin, to address participant concerns and misconceptions about insulin, and to provide positive messages about insulin and the importance of good glucose control (19,20). As GRADE was not designed for assessment of the efficacy of strategies to promote rescue insulin use, the impact of these strategies on rescue insulin use is not known.
Delayed intensification of diabetes treatment, often until A1C values are well in excess of recommended targets, is a well-recognized problem, and patient and provider beliefs about the necessity for intensification, and the potential risks and burdens of intensification, are known contributing factors (8,15,17,18,21–24). Presumably, investigators agreed with the GRADE protocol for rescue insulin, and participants likewise agreed as indicated by their written consent. Yet, as we have shown, adherence to the rescue insulin protocol was low relative to expectations. While the letter of the protocol for rescue insulin was explicit, application of the protocol appears to have been more nuanced, perhaps reflecting a more individualized approach to diabetes treatment, as would be appropriate in clinical care. Our finding that higher secondary outcome A1C values were associated with rescue insulin use (viewed differently, lower A1Cs were associated with nonuse) may be an indicator of this. For example, A1C values just over 7.5% (58 mmol/mol) may have been viewed as amenable to improvement (thus, averting the need for rescue insulin) by changes to dietary or exercise practices, by addressing issues of adherence to the current prescribed treatment, or by further optimizing the dose of the assigned treatment (e.g., continued titration of glargine). These were common themes in the NONINIT survey results (Fig. 2).
The association we found of older age with not using rescue insulin is consistent with other reports (15,17,23). Although current diabetes treatment guidelines support relaxing glycemic goals or de-intensifying treatment for some elderly patients, with multiple comorbidities or limited life expectancy, it is unlikely that any of these factors had a significant impact on rescue insulin use in GRADE (25–28). Mean ± SD age of GRADE participants at baseline was 57.2 ± 10.0 years, and those with significant comorbidities at baseline were not eligible for the study (11). While health status or life circumstances (for example, taking a job with insulin use restrictions) may have changed during GRADE, none of the NONINIT survey reasons addressing the appropriateness of an A1C target of ≤7.5% (≤58 mmol/mol) or prohibitions on insulin use were among those commonly reported.
The NONINIT survey reasons “Making diet/lifestyle changes” and “Intention to improve adherence to treatment” were among the most commonly reported, mirroring findings from the Translating Research Into Action for Diabetes (TRIAD) Study Insulin Starts Project, in which patient intention to change health behaviors was the primary reason for not starting insulin as prescribed (29). The NONINIT reason “Does not want to take insulin” was more common than “Not wanting to take additional medications” or “Fear of injections.” Reluctance to use insulin per se has also been reported, suggesting that more needs to be done to destigmatize insulin as a treatment for type 2 diabetes (5,9,29). The need to keep GRADE participants (all research volunteers were free to withdraw from participation at any time) actively engaged in GRADE to its conclusion might have affected how aggressively investigators pursued rescue insulin among participants expressing reluctance. However, the NONINIT reason “Participant will likely not continue in study due to concerns about taking insulin” was not among the commonly recorded reasons. While not specifically addressed in the NONINIT survey, provider or participant research fatigue in relation to failing to implement rescue insulin cannot be dismissed, given the duration and intensity of follow-up in GRADE (30).
Strengths of this study include the large sample size, racial and ethnic diversity of the participants, wide geographic distribution of clinical centers, careful characterization of participants, and longitudinal measurement of glycemia, weight, hypoglycemia (albeit by self-report only), and other outcomes over a mean follow-up of 5 years. Despite the COVID-19 pandemic during the last 18 months of GRADE, implementation of remote and limited in-person study visits, use of home capillary A1C collections (31), and gradual lifting of prohibitions of in-person study visits resulted in completion of 89% of all expected visits during the final year of the study, and 94% of participants completed the final study visit (11).
Our study has several limitations. Participants were carefully selected to meet the prescribing criteria for all four treatments, an A1C target of <7% (<53 mmol/mol) had to be an appropriate treatment goal, and participants had to be willing to accept random assignment and agree to intensify treatment with insulin. Therefore, the cohort was already predisposed to accept insulin. These factors limit the generalizability of our findings to less highly selected populations.
The NONINIT survey results that we have included are meant to provide context and insights into our findings; however, the conclusions we draw from these results are speculative and must be considered with caution. The survey was developed for internal study monitoring purposes in response to higher-than-expected numbers of rescue insulin use protocol deviations. While many of the reasons for non-initiation of insulin included in the survey are supported by the literature, they were developed through discussion and consensus and reflect the clinical experiences and beliefs of the GRADE Research Group members, most notably the physicians, nurses, and diabetes educators who served on the Protocol Oversight Committee. Surveys were completed by the study team members based on their interactions with the participants, not by the participants themselves, which may introduce bias in the selection and ranking of the reasons. Finally, while the NONINIT survey allowed the selection of up to four reasons for not using rescue insulin, this analysis only included the reasons listed as primary. Therefore, these results likely do not capture the complexities of the treatment decision-making process, which might be better revealed in a more in-depth analysis of the NONINIT results.
Although treatment intensification in GRADE was limited to rescue glargine and aspart, GRADE investigators were permitted to refer participants to non-GRADE providers for alternative, noninsulin treatments and were required to do so for participants with known cardiovascular or kidney disease following the publication of consensus recommendations for the preferential use of GLP-1 receptor agonists and sodium–glucose cotransporter 2 inhibitors in these populations (27,32). A modest fraction of the secondary and tertiary outcome sample (5% and 10%, respectively) was known to have started a non-GRADE glucose-lowering medication prior to outcome confirmation. However, in a sensitivity analysis where we repeated the longitudinal analysis of A1C change following secondary and tertiary outcome but excluded data from any visit after the introduction of a nonstudy glucose-lowering agent, the results presented were not altered.
In conclusion, treatment regimens for individuals with type 2 diabetes frequently must be intensified to achieve and maintain recommended levels of glycemia. Failing to change or intensify treatment when required is a recognized and persistent problem in diabetes care. We have shown that even in the setting of a well-resourced clinical study with carefully selected participants and skilled diabetes care providers, intensifying treatment using insulin proved challenging.
The availability of newer drugs such as sodium–glucose cotransporter 2 inhibitors and especially GLP-1 receptor agonists provides more noninsulin options for many patients with type 2 diabetes, and their use is well outlined in recent guidelines (27,33). Insulin, however, continues to be a major and too often delayed option for treating type 2 diabetes. The potential availability of a weekly basal insulin, and strategies like continuous glucose monitoring, especially for those fearful of, or at heightened risk for, hypoglycemia, could help persuade providers to prescribe insulin and help individuals with type 2 diabetes accept insulin treatment (34). Finally, while it is reassuring that many GRADE participants had improved glycemia by study end (including a considerable proportion who never added rescue insulin), the current results, and findings from GRADE more generally (11), support the need for more widely acceptable and effective strategies to help people with type 2 diabetes achieve and maintain lasting glucose control.
Clinical trial reg. no. NCT01794143, clinicaltrials.gov
This article contains supplementary material online at https://doi.org/10.2337/figshare.23800437.
A complete list of members of the GRADE Research Group can be found in the supplementary material online.
This article is featured in podcasts available at diabetesjournals.org/care/pages/diabetes_care_on_air.
Article Information
Acknowledgments. The GRADE Research Group is deeply grateful to the participants of GRADE, whose loyal dedication made GRADE possible.
Funding. GRADE was supported by a grant from the NIDDK of the National Institutes of Health under award no. U01DK098246. The planning of GRADE was supported by a U34 planning grant from the NIDDK (U34-DK-088043). The American Diabetes Association supported the initial planning meeting for the U34 proposal. The National Heart, Lung, and Blood Institute and the Centers for Disease Control and Prevention also provided funding support. The Department of Veterans Affairs provided resources and facilities. Additional support was provided by National Institutes of Health grants P30 DK017047, P30 DK020541, P30 DK020572, P30 DK072476, P30 DK079626, P30 DK092926, U54 GM104940, UL1 TR000170, UL1 TR000439, UL1 TR000445, UL1 TR001102, UL1 TR001108, UL1 TR001409, 2UL1TR001425, UL1 TR001449, UL1 TR002243, UL1 TR002345, UL1 TR002378, UL1 TR002489, UL1 TR002529, UL1 TR002535, UL1 TR002537, UL1 TR002541, and UL1 TR002548. Educational materials were provided by the National Diabetes Education Program. Material support in the form of donated medications and supplies was provided by Becton, Dickinson and Company, Bristol-Myers Squibb, Merck & Co., Novo Nordisk, Roche Diagnostics, and Sanofi.
The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Duality of Interest. A.J.A. reports personal fees from Medtronic outside the submitted work. M.L.J. reports grants from Eli Lilly, Novo Nordisk, and Sanofi outside the submitted work. N.R. and H.J.W. report grants from Novo Nordisk and SomaLogic, and other support from Novo Nordisk and Eli Lilly outside the submitted work. No other potential conflicts of interest relevant to this article were reported.
Author Contributions. All authors affirmed that authorship is merited based on the International Committee of Medical Journal Editors (ICMJE) authorship criteria. P.A.H., H.K.-S., N.M.B., M.L.J., T.K., N.R., and C.L.M. contributed to the conception and design of the research. H.K.-S., A.J.A., B.N.F., M.L.J., T.K., N.R., H.J.W., and C.L.M. contributed to acquisition of data. H.K.-S., N.M.B., and E.J.K. contributed to statistical analysis of data. P.A.H., H.K.-S., N.M.B., E.J.K., A.J.A., B.N.F., M.L.J., T.K., M.E.L., E.A.L., N.R., H.J.W., and C.L.M. contributed to interpretation of data and results. H.K.-S. contributed to acquisition of funding. P.A.H., H.K.-S., B.N.F., M.L.J., T.K., V.S.L., E.A.L., N.R., and C.L.M. contributed to supervision and management of research. P.A.H., H.K.-S., N.M.B., E.J.K., and C.L.M. contributed to drafting of the manuscript. H.K.-S., N.M.B., E.J.K., A.J.A., B.N.F., M.L.J., T.K., V.S.L., M.E.L., E.A.L., N.R., and H.J.W. contributed to critical revision of the manuscript. P.A.H. and H.K.-S. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Prior Presentation. Parts of this study were presented in abstract form at the 82nd Scientific Sessions of the American Diabetes Association, New Orleans, LA, 3–7 June 2022.