Factors associated with A1C reduction with GLP-1 agonist or SGLT-2 inhibitor use

Abstract

Background

While use of glucagon-like peptide-1 (GLP-1) agonists and sodium-glucose cotransporter-2 (SGLT-2) inhibitors reduces the risk of atherosclerotic cardiovascular disease outcomes and lowers glycosylated haemoglobin (A1C), evidence on patient characteristics associated with clinically relevant A1C reduction is lacking.

Objective

The objective of this retrospective cohort study was to identify patient characteristics associated with A1C reduction with initial GLP-1 or SGLT-2 use.

Methods

Patients with type 2 diabetes and a baseline A1C ≥7% who were dispensed a GLP-1 or SGLT-2 between 01/01/10 and 12/31/17 were included. Patients were categorized as having a ≥1% or <1% A1C reduction during the 90–365 days after GLP-1/SGLT-2 initiation. Patient characteristics were collected during the 180 days prior to initiation. Multivariable logistic and linear regression modelling was performed to identify characteristics associated with a ≥1% A1C reduction and absolute change in A1C, respectively.

Results

Five hundred and seventy-two patients were included with 261 (46%) and 311 (54%) having and not having an ≥1% A1C reduction. Patients were primarily middle-aged, female, white, non-Hispanic and had a high burden of chronic disease. Characteristics associated with a ≥1% A1C reduction included: GLP-1/SGLT-2 persistence, congestive heart failure comorbidity, phentermine dispensing, care management team (CMT) enrollee and higher baseline A1C. Characteristics associated with absolute A1C reduction included: age, baseline A1C, CMT enrollee, GLP-1/SGLT-2 persistence and a phentermine dispensing.

Conclusions

The results of this study provide practitioners with guidance on the patients who are most likely to have a clinically relevant A1C reduction with GLP-1 or SGLT-2 use.

Diabetes mellitus, glucagon-like peptide-1 receptor, glycated haemoglobin A, patient care management, regression analysis, sodium-glucose transporter-2 inhibitors

Key Messages

• Patient factors associated with A1C reduction were identified.

• Identified factors included GLP-1/SGLT-2 persistence, congestive heart failure and phentermine use.

• Other factors included higher baseline A1C and care management team enrollee.

Introduction

The American Diabetes Association (ADA) 2020 Guidelines recommend that after metformin therapy, there are six medication classes that can be used as dual therapy or triple therapy for further type 2 diabetes (T2DM) treatment (1). A patient-centred approach that takes into consideration patient-specific factors impacting choice of therapy is recommended. Key factors include comorbidities, individualized glycosylated haemoglobin (A1C) target, risk of hypoglycaemia and weight as well as socioeconomic factors (1). Among those patients with T2DM and established atherosclerotic cardiovascular disease (ASCVD), a glucagon-like peptide-1 receptor agonist (GLP-1) or sodium-glucose cotransporter-2 inhibitor (SGLT-2) is recommended as these agents have been shown to reduce the risk of major adverse cardiovascular events (MACEs) (1,2). In addition, SGLT-2 is recommended for patients with existing heart failure or chronic kidney disease while both agents were recommended to minimize weight gain or promote weight loss (1,2).

Evidence supports the use of GLP-1 and SGLT-2 in certain patient populations; however, their macro- and microvasculature outcomes appear to be independent of A1C values. Nevertheless, measurement of A1C remains the gold standard for diagnosing and assessment of diabetes control for medical therapy management (3). The Diabetes Control and Complications Trial (DCCT) and the United Kingdom Prospective Diabetes Study (UKPDS) emphasize that maintaining glycaemic control is necessary to reduce the risk of microvascular and macrovascular complications (4,5). An epidemiological review of the UKPDS revealed that a reduction in A1C of at least 1% resulted in a 35% reduction in the microvascular complications (i.e. retinopathy, nephropathy and neuropathy) in T2DM (4,5).

The GLP-1 and SGLT-2 have demonstrated mean reductions in A1C ranging from 0.7% to 1.7% and 0.3% to 1.2%, respectively (6). Unfortunately, there is limited evidence describing patient characteristics that are associated with a clinically relevant A1C reduction with use of either medication class, particularly relevant for those patients without aforementioned comorbidities.

Previous studies suggest that for GLP-1, higher baseline A1C, previous treatment with metformin, longer duration of T2DM and mean prandial blood glucose levels are associated with A1C reduction (7–9). For SGLT-2, higher baseline A1C, preserved estimated glomerular filtration rate, male sex and baseline c-peptide index have been associated with A1C reduction (10,11). In addition to variation between results, these studies are limited by small sample sizes, varying study designs and an evaluation of laboratory markers that are not routinely measured in clinical practice (7–11).

It remains unclear which patient characteristics are associated with a meaningful A1C reduction with either GLP-1 or SGLT-2 use. The aim of this study was to determine characteristics associated with an A1C reduction of ≥1% and absolute change in A1C in patients who received a GLP-1 or SGLT-2 within an integrated health care delivery system. Identification of such characteristics will further guide use in clinical practice and, thus, avoid unnecessary exposure and costs to patients and health system organizations.

Materials and methods

Study design and setting

This was a retrospective cohort study of patients with T2DM who received either a GLP-1 or SGLT-2 between 1/1/2010 and 12/31/2017. The study was conducted at Kaiser Permanente Colorado (KPCO), an integrated health care delivery system providing care to approximately 660 000 members in Colorado. At KPCO, GLP-1 and SGLT-2 are non-preferred for patients without a compelling indication and use is restricted to patients who are unable to achieve glycaemic targets with preferred anti-diabetic medications.

KPCO has its own pharmacies where patients can receive subsidized prescription medications. Coded and free-text medical, pharmacy, laboratory, emergency department, hospitalization and membership information from within the integrated health care delivery system, as well as from other contracted and affiliated facilities, are captured in KPCO’s administrative and claims databases.

The date of a patient’s initial GLP-1/SGLT-2 dispensing during the study period was set as the index date. Patients had no other GLP-1/SGLT-2 dispensing prior to the index date during their membership at KPCO. Utilizing the A1C measurement recorded prior to and most proximal to the index date and the earliest A1C measurement between 90 and 365 days after the index date, patients were categorized based on having had an A1C reduction ≥1% from baseline to the follow-up A1C or not. A follow-up period of at least 90 days after the index date was chosen as the A1C test measures the previous 90 days’ average glycosylated haemoglobin. A limit of 365 days was chosen as medication effectiveness is likely to be re-assessed after this time (2).

Patient population

Patients who were KPCO members, ≥18 years of age as of the index date and diagnosed with T2DM were eligible for the study. Patients who had their initial GLP-1 (i.e. exenatide, liraglutide, lixisenatide, albiglutide, dulaglutide, semaglutide) or SGLT-2 (i.e. canagliflozin, dapagliflozin, empagliflozin, ertugliflozin, empagliflozin + linagliptin, empagliflozin + metformin, dapagliflozin + metformin) dispensed during the study period, an A1C ≥7% recorded within 180 days prior to index date (baseline), and an A1C recorded during the 90–365 days after the index date (follow-up) were included. Patients were excluded if they had a diagnosis of gestational diabetes or type 1 diabetes or a positive human chorionic gonadotropin test recorded during baseline or between the index date and 365 days after the index date.

Study outcomes

The primary outcome was patient characteristics associated with an A1C reduction ≥1% after initiation of a GLP-1 or SGLT-2. Secondary outcomes included comparisons of patient characteristics between those who did (responders) and did not (non-responders) have an A1C reduction ≥1%. An A1C reduction ≥1% was chosen since it has been associated with decreased micro- and macrovascular complications (4,5,12). In addition, the absolute change in A1C after initiation of a GLP-1 or SGLT-2 was determined along with patient characteristics that were associated independently with absolute change.

Data collection

Patient characteristics were identified from electronic queries of the KPCO electronic laboratory, medical, membership, pharmacy and procedure databases. Information on GLP-1 and SGLT-2 dispensings were obtained from electronic queries of the administrative pharmacy database using National Drug Code numbers. Patient medical record numbers obtained from these queries were used to query other databases.

Characteristics for assessment of association were identified based on clinical judgement, in consideration of routine diabetes monitoring parameters, statistical (i.e. P < 0.2 in bivariate analysis) and information from the literature (7–9,11). Information on comorbidities, demographics, laboratory measures and procedures were obtained during baseline (i.e. during the 180 days prior to the index date). In addition, information on baseline dispensings of medications that influenced weight loss (e.g. phentermine and topiramate) were obtained. Laboratory measures for the A1C reduction outcome were obtained during follow-up. Samples of extracted data were verified for accuracy by manual chart review.

Data analyses

Change in A1C was determined by subtracting the baseline A1C from the first A1C recorded during the follow-up. Patients were categorized as responders or non-responders based on absolute change in A1C. Age was calculated as of the index date. Race was categorized as black (African American), white, American Indian or Alaska Native, Asian, Native Hawaiian or Other Pacific Islander or Undeclared. Care management team (CMT) enrollee was defined as the patient having received dedicated care, as of the index date, from either a clinical pharmacist or registered nurse who provided guidance for shared decision-making, as well as the selection and sequence of appropriate pharmacological treatment for diabetes (13). Persistence with GLP-1/SGLT-2 therapy was defined as a days supply dispensed that encompassed the date of the follow-up A1C laboratory measurement. Index GLP-1/SGLT-2 prescriber department was categorized as endocrinology, primary care, weight management/bariatric surgery or other/unknown. Weight was categorized as ≤215.5 lbs, >215.5 to ≤260.1 lbs, >260.1 lbs or missing.

A chronic disease score, a validated measure of a patient’s burden of chronic illness, was calculated for each patient using ambulatory prescription medication dispensings during baseline (14). The chronic disease score ranges in value from 0 to 36, with increasing values indicating a higher burden of chronic illness. It provides an adjusted risk of hospitalization and resource use. The presence of specific comorbidities was determined using the Quan adaptation of the Charlson comorbidity index (15). The algorithm was applied to diagnoses that were recorded during baseline to provide a 0–30 comorbidity score that provides an adjusted risk of mortality and resource use.

Patient characteristics were summarized using descriptive statistics including means (±standard deviation) for continuous variables and percentages for categorical variables. Characteristics were compared between responder and non-responder groups in univariate analyses. T-tests were used to compare continuous variables and chi-square tests of association or Fisher’s exact tests were used to compare categorical variable. A multivariable logistic regression model was constructed to determine patient characteristics independently associated with an A1C reduction ≥1% A1C. A c-statistic was calculated to assess model discriminability. A multivariable linear regression model was constructed to determine patient characteristics associated with absolute change in A1C. All patient characteristics (except for laboratory and blood pressure values as a significant number of patients were missing these values) were included in the model. All characteristics were entered and then backward eliminated manually until only characteristics with a P value <0.05 remained. Analyses were performed using SAS v9.4 (SAS Institute Inc., Cary, NC) and the alpha was set at 0.05.

Results

A total of 572 patients were included with 261 (46%) and 311 (54%) responders and non-responders, respectively (Fig. 1). Overall, patients were primarily middle-aged, female, white, non-Hispanic, had received a GLP-1 and had a high burden of chronic disease (Table 1). A total of 38 patient characteristics were assessed. Patient characteristics were similar between groups; however, responders were more likely to have been enrolled in CMT and persistent with index GLP-1/SGLT-2 and had a higher baseline A1C and an outpatient prescription for phentermine dispensed (Table 1).

Table 1.

Univariate comparisons of patient characteristics by A1C response status in 582 patients from an integrated health care delivery system who received either a GLP-1 or SGLT-2 between 1/1/2010 and 12/31/2017^a

	Responder (n = 261)	Non-responder (n = 311)	P value
Demographic
Mean age (years, SD)	56.0 (10.5)	57.0 (10.7)	0.260
Female sex (n, %)	150, 57.5%	171, 55.0%	0.550
Race (n, %)			0.812
American Indian/Alaska Native	3, 1.2%	7, 2.3%
Asian	4, 1.5%	5, 1.6%
Black/African American	18, 6.9%	15, 4.8%
White	165, 63.2%	197, 63.3%
Native Hawaiian/Other Pacific Islander	12, 4.6%	17, 5.5%
Undeclared	59, 22.6%	70, 22.5%
Hispanic ethnicity (n, %)	66, 25.3%	67, 21.5%	0.291
CMT enrollee (n, %)	123, 47.1%	93, 29.9%	<0.001
High deductible health plan (n, %)	7, 2.3%	8, 2.6%	0.651
Medicare beneficiary (n, %)	82, 31.4%	102, 32.8%	0.632
Medicaid beneficiary (n, %)	37, 14.2%	36, 11.6%	0.406
Index prescription information
GLP-1 index medication dispensing (n, %)	226, 86.6%	252, 81.0%	0.074
Persistent with SGLT-2/GLP-1 at date of follow-up A1C measure	167, 64.0%	136, 43.7%	<0.001
Physician prescriber (n, %)	203, 77.8%	230, 74.0%	0.288
Prescriber department (n, %)			0.299
Endocrinology	67, 25.7%	101, 32.5%
Primary care	96, 36.8%	105, 33.8%
Weight management/bariatric surgery	30, 11.5%	37, 11.9%
Other/unknown	68, 26.1%	68, 21.9%
Laboratory measure
Mean A1C (SD)	9.9 (1.7)	8.9 (1.3)	<0.001
Mean creatinine (n, SD)	0.9 (200, 0.3)	0.9 (250, 0.3)	0.404
Mean glomerular filtration rate (n, SD)	80.7 (187, 23.0)	82.8 (222, 27.8)	0.957
Mean low-density lipoprotein cholesterol (n, SD)	82.5 (81, 29.8)	74.9 (112, 29.1)	0.168
Vital measure
Mean baseline weight (pounds, n, SD)	248.2 (218, 60.2)	238.8 (236, 51.9)	0.707
Weight categories (pounds, n, %)			0.124
≤215.5	69, 26.4%	81, 26.1%
>215.5 to ≤260.1	71, 27.2%	79, 25.4%
>260.1	78, 29.9%	76, 24.4%
Missing	43, 16.5%	75, 24.1%
Mean diastolic blood pressure (mmHg, n, SD)	73.5 (233, 10.0)	75.0 (263, 10.5)	0.198
Mean systolic blood pressure (mmHg, n, SD)	127.9 (233, 16.2)	130.3 (263, 15.4)	0.193
Outpatient medication dispensing
ACE/ARB (n, %)	135, 51.7%	171, 55.0%	0.436
Insulin (n, %)	162, 62.1%	173, 55.6%	0.119
Metformin (n, %)	199, 76.3%	226, 72.7%	0.330
Phentermine (n, %)	46, 17.6%	32, 10.3%	0.011
Pioglitazone (n, %)	16, 6.1%	17, 5.5%	0.734
Sulfonylurea (n, %)	129, 49.4%	133, 42.8%	0.111
Statin (n, %)	134, 51.3%	155, 49.8%	0.721
Topiramate (n, %)	7, 2.7%	8, 2.6%	0.935
Comorbidity
Cerebrovascular disease (n, %)	7, 2.3%	5, 1.6%	0.372
Chronic pulmonary disease (n, %)	54, 20.7%	47, 15.1%	0.081
Congestive heart failure (n, %)	21, 8.1%	12, 3.9%	0.032
Diabetes with chronic complications (n, %)	116, 44.4%	137, 44.1%	0.925
Renal insufficiency (n, %)	41, 15.7%	63, 20.3%	0.160
Coronary artery disease
Coronary artery bypass grafting (n, %)	1, 0.4%	3, 1.0%	0.406
MI (n, %)	6, 2.3%	18, 5.8%	0.038
Percutaneous coronary intervention ± stenting (n, %)	2, 0.8%	2, 0.6%	0.860
Risk scores
Mean Charlson comorbidity index (SD)	2.4 (1.9)	2.3 (1.6)	0.381
Mean chronic disease score (SD)	6.4 (2.8)	6.1 (3.3)	0.251

^aPatients in the response group had an A1C reduction of ≥1% from baseline.

Figure 1.

Dispositions of patients from an integrated health care delivery system who received either a GLP-1 or SGLT-2 between 1/1/2010 and 12/31/2017.

Characteristics included in the logistic regression model were age, baseline A1C, chronic pulmonary disease (CPD) comorbidity, congestive heart failure (CHF) comorbidity, CMT enrollee, female sex, GLP-1 or SGLT-2 index medication, insulin dispensing, myocardial infarction (MI) comorbidity, persistence with GLP-1/SGLT-2 index medication, phentermine dispensing, renal insufficiency comorbidity, sulfonylurea dispensing and weight category (Table 2). Patient characteristics associated with an increased likelihood of an A1C reduction ≥1% included a higher baseline A1C, CHF, CMT enrollee, GLP-1/SGLT-2 persistence and phentermine dispensing. A MI was associated with a decreased likelihood of an A1C reduction ≥1%.

Table 2.

Results of multivariate logistic regression modelling of the association of patient characteristics with a ≥1% absolute reduction in A1C among 512 patients from an integrated health care delivery system who received either a GLP-1 or SGLT-2 between 1/1/2010 and 12/31/2017

Characteristic	Odds ratio	95% confidence interval
Age	1.02	0.99–1.04
Baseline A1C value	1.81	1.56–2.09
Chronic pulmonary disease comorbidity	1.42	0.86–2.35
CHF comorbidity	2.76	1.13–6.75
CMT enrollee	1.86	1.22–2.84
Female sex	1.05	0.69–1.59
GLP-1 index medication	1.15	0.67–1.99
Baseline insulin dispensing	0.93	0.61–1.41
MI comorbidity	0.23	0.08–0.68
Persistent with GLP-1/SGLT-2	2.92	1.98–4.33
Baseline phentermine dispensing	2.12	1.16–3.85
Renal insufficiency comorbidity	0.60	0.35–1.02
Baseline sulfonylurea dispensing	1.22	0.81–1.82
Baseline weight category
≤215.5 pounds	1.21	0.67–2.17
>215.5 to ≤260.1 pounds	1.27	0.71–2.26
>260.1 pounds	1.25	0.69–2.24
Missing (referent)	—	—

c-Statistic = 0.779.

Overall, the mean change in A1C was −1.0 ± 1.8%. Characteristics included in the linear regression model were the same as the logistic regression and angiotensin-converting enzyme inhibitor (ACE)/angiotensin II receptor blocker (ARB) dispensing, cerebrovascular disease comorbidity, Charlson comorbidity index, chronic disease score, high deductible health plan, Hispanic ethnicity, length of time (in days) between baseline and follow-up A1C measurement, Medicaid beneficiary, Medicare beneficiary, metformin dispensing, physician prescriber, pioglitazone dispensing, prescriber department, race, sulfonylurea dispensing, statin dispensing, topiramate dispensing and weight category. After backward elimination, remaining in the model included ACE/ARB dispensing, age, baseline A1C, CMT, persistence with GLP-1/SGLT-2 index medication, phentermine dispensing, MI and renal insufficiency comorbidity. The most impactful of these characteristics on absolute decrease and increase in A1C was persistence with GLP-1/SGLT-2 index medication and MI, respectively (Table 3).

Table 3.

Results of multivariate linear regression modelling of patient characteristics associated with absolute A1C change with backward elimination of factors with a P value <0.05 among 512 patients from an integrated health care delivery system who received either a GLP-1 or SGLT-2 between 1/1/2010 and 12/31/2017

Characteristic	Β coefficient	Standard error	P value
Intercept	5.253	0.589	<0.001
Baseline ACE/ARB dispensing	0.275	0.130	0.035
Baseline age	−0.020	0.007	0.003
Baseline A1C	−0.503	0.041	<0.001
CMT enrollee	−0.290	0.133	0.029
MI comorbidity	0.721	0.337	0.030
Persistent with GLP-1/SGLT-2	−0.847	0.127	<0.001
Baseline phentermine dispensing	−0.544	0.192	0.005
Renal insufficiency comorbidity	0.508	0.171	0.003

Adjusted R-square = 0.282.

Discussion

While GLP-1 agonists and SGLT-2 inhibitors are recommended for patients with a history of ASCVD to prevent MACE, they also improve glycaemic response. Yet, little real-world evidence exists in the USA of the patient characteristics associated with clinically relevant improvements in glycaemic response with their use. This naturalistic study of 572 patients with T2DM who received a GLP-1 or SGLT-2 from an integrated health care delivery system identified five patient characteristics that were associated independently with an increased likelihood of a clinically relevant A1C reduction and eight patient characteristics associated independently with absolute change in A1C. The characteristics we identified as most strongly associated with an A1C reduction ≥1% included GLP-1/SGLT-2 persistence (i.e. patients who were persistent were 2.9× more likely to achieve the reduction) and CHF (i.e. patients with a CHF comorbidity were 2.8× more likely to achieve the reduction). The characteristic we identified as most strongly associated with an absolute decrease in A1C was GLP-1/SGLT-2 persistence (i.e. patients who were persistent were associated with a −0.85% absolute decrease in A1C). Additionally, a phentermine dispensing, being a CMT enrollee, and a higher baseline A1C were associated with an A1C reduction ≥1%. Age was associated with a small absolute A1C decrease. Only a MI was associated with a decreased likelihood of an A1C reduction ≥1% (i.e. patients with a history of a MI had a 77% lower likelihood of achieving a reduction). An ACE/ARB dispensing, MI and renal insufficiency comorbidity were associated with an absolute increase in A1C. These results are important as they add to the small body of evidence evaluating which patients derive the greatest A1C reducing benefit from use of a GLP-1 or SGLT-2.

In one small, naturalistic study of A1C response to GLP-1 initiation, Khan et al. assessed ‘predictors’ of 11 mmol/mol (≈1%) reduction in A1C after 6 months of therapy in 112 patients with T2DM (7). They identified that higher baseline A1C was the sole predictor of response (7). In an even smaller study, Imai et al. assessed clinical parameters associated with achievement of an A1C <8% (response) in 43 hospitalized patients treated with a GLP-1 (8). They identified that the anti-diabetes treatment used before the initiation of therapy, body mass index and pre-prandial blood glucose level during the 2 days after the initiation of therapy were associated with response (8). Rosenstock et al. performed a subgroup analysis of a Phase III study of 259 patients with T2DM where a GLP-1 or placebo was added to insulin therapy (9). They reported that longer diabetes duration and lower BMI were associated with greater A1C reductions (9). Abe et al. in their pooled analysis of 932 patients from tofogliflozin (SGLT-2) Phase II and III studies identified that higher baseline A1C, higher estimated glomerular filtration rate, male sex and shorter duration of T2DM were associated with a reduction in A1C (11). Similarly, Cho et al. examined the medical records of 374 patients with T2DM who were initiated on a SGLT-2 to identify clinical parameters associated with a glycaemic response (16). They reported that younger age, higher baseline A1C, higher estimated glomerular filtration rate and shorter duration of T2DM were associated with glycaemic response (16). We identified no published studies that included patients who received either a GLP-1 or SGLT-2 and assessed glycaemic response.

Our finding of a higher baseline A1C being associated with an increased likelihood of response is consistent with previous literature. It is intuitive that a patient with a higher A1C can achieve more readily an absolute 1% reduction than a patient with a lower A1C. Contrary to previous studies, however, we did not find an association with baseline anti-diabetes treatment, sex, kidney function or BMI (weight). We did not assess duration of diabetes. The differences in our and other studies’ findings may be explained by variations in study designs (9,11), smaller sample sizes (7–9) and the exclusive use of only one medication class (1,7,9,11).

This is the first study to identify medication persistence, phentermine use and CHF as independent factors associated with glycaemic response. In addition, we identified that enrolment in a diabetes care management program was associated with glycaemic response. It is logical that medication persistence would be associated with improved glycaemic response (17) and disease management from a care team dedicated to the core elements of diabetes chronic care should be associated with improved glycaemic response (18). For patients with T2DM, weight loss of at least 5% can produce beneficial outcomes in glycaemic response (19). Nevertheless, two small trials evaluating phentermine for weight loss in patients with T2DM did not find an effect on fasting blood glucose (20,21).

It is unclear why our finding that a MI was associated with a decreased likelihood of response. We had relatively few patients with this diagnosis, which may have led to an unrepresentative sample. For the most part, patients were initiated on a GLP-1/SGLT-2 prior to widespread evidence of the cardiovascular benefits of these medication classes. Though a CHF comorbidity resulted in a high likelihood of response, the mechanism of this relationship is unclear. Nevertheless, it is an intriguing finding considering the ADA recommends an SGLT-2 in patients with T2DM and CHF, independent of A1C impact (1) and new data suggest beneficial response on CHF outcome without diabetes (22,23). GLP-1 agonists have not been found to be associated independently with improved heart failure outcomes (1) but were likely responsible for driving the outcomes in this study due to our large proportional sample of patients with this exposure.

Limitations

While our study provided real-world insights regarding patient characteristics and clinically meaningful glycaemic response to GLP-1 and SGLT-2, it had several limitations. We were unable to perform separate analyses for the patients who received a GLP-1 versus SGLT-2 because of the small number of patients dispensed an SGLT-2 during the study period. This might render the results less generalizable to patients using an SGLT-2 (24). We did not assess long-term A1C response. We only assessed GLP-1 and SGLT-2 naïve patients and did not examine patients who switched between these therapies. Finally, the results of this study were derived from one integrated health care delivery system. Results may be different in other study settings (e.g. one without a dedicated diabetes CMT).

Conclusions

This retrospective analysis of real-world patients identified that a reduction in A1C ≥1% with use of a GLP-1 agonist or SGLT-2 inhibitor was associated with three novel patient characteristics (medication persistence, phentermine use, CHF) while only one (MI) was associated with decreased likelihood of response. In addition, we found that an ACE/ARB dispensing, MI and renal insufficiency comorbidity were associated with an increase in absolute A1C. These results provide practitioners with additional guidance regarding patients who are most likely to have a clinically relevant A1C reduction with GLP-1 or SGLT-2 use.

Declaration

Funding: this study was funded by Kaiser Permanente. The funder had no role in the study design, collection, analysis and interpretation of data, writing of the report or the decision to submit the manuscript for publication.

Ethical approval: all aspects of this study were reviewed and approved by the Kaiser Permanente Colorado Institutional Review Board.

Conflict of interest: the authors have no conflicts of interest to report. Preliminary results were presented at the Mountain States Conference for Residents and Preceptors in Salt Lake City, UT on 9 May 2019.