Impact of GAD-65 antibody detection on antihyperglycemic therapy in T2D

Abstract

Guidelines recommend that latent autoimmune diabetes in adults be managed similarly to type 1 diabetes, emphasizing early insulin initiation, while other guidance suggests incorporating non-insulin antihyperglycemic therapies based on comorbidities and C-peptide levels. The objective of this study was to explore whether glutamic acid decarboxylase-65 autoantibody detection in individuals diagnosed with type 2 diabetes impacts antihyperglycemic treatment. This retrospective, observational cohort study at an academic integrated delivery network in Northeast Ohio utilized electronic medical records. Individuals previously diagnosed with type 2 diabetes who tested positive for glutamic acid decarboxylase-65 autoantibodies between September 1, 2021, and September 1, 2023, were included. The primary outcome compared the median number of non-insulin antihyperglycemic medication classes at the time of autoantibody detection and after 12 months. Participants had a mean age of 57.2 ± 13.7 years, 73.3% were white, with a median body mass index of 26.2 kg/m², and 82.1% had an estimated glomerular filtration rate greater than 60 mL/min/1.73 m². A median of 1 (0–2) non-insulin antihyperglycemic medication class was present at detection and 1 (0–1.5) post-detection (P < 0.0001). Significantly fewer individuals were prescribed metformin, dipeptidyl peptidase-4 inhibitors, glucagon-like peptide-1 receptor agonists, sodium-glucose cotransporter 2 inhibitors, sulfonylureas, and thiazolidinediones, whereas significantly more individuals were prescribed bolus insulin after detection. Although most individuals had a baseline C-peptide value within 12 months prior to detection, significantly fewer C-peptide assessments occurred during the subsequent 12-month follow-up period. In individuals with type 2 diabetes, glutamic acid decarboxylase-65 autoantibody detection was associated with significantly fewer prescribed non-insulin antihyperglycemics and significantly more insulin agents 12 months after detection.

Introduction

Latent autoimmune diabetes in adults (LADA) is a chronic endocrine disorder characterized by impaired glucose regulation via autoimmune-mediated destruction of pancreatic beta cells (1, 2). Clinical presentation of LADA at diagnosis can resemble type 2 diabetes, whereas the immunologic pathophysiology more closely resembles type 1 diabetes, coining the nickname ‘type 1.5 diabetes’ (2, 3, 4). Core LADA diagnostic criteria include onset at 30 years of age or older, seropositive diabetes-associated autoantibodies, and the absence of insulin requirements at 6 months after diagnosis (2). Among associated autoantibodies, glutamic acid decarboxylase-65 (GAD-65) is classified as the most sensitive marker for LADA diagnosis, where approximately 90% of participants with LADA in the Action LADA 7 study had GAD-65 autoantibody (AAb) detected (2, 5). The onset of LADA often occurs later in life due to the slow progression of immunogenicity, accounting for 2–12% of all adult-onset diabetes diagnoses (2, 3). Given its autoimmune-mediated mechanism, individuals with LADA often have a personal or family history of autoimmunity but less frequently have metabolic syndrome compared with individuals with type 2 diabetes (2). Accordingly, body mass index (BMI) and blood pressure trend lower and high-density lipoprotein (HDL) cholesterol trends within reference ranges in LADA populations compared with those with type 2 diabetes (2). Autoantibody testing may be clinically indicated in patients with abrupt glycemic deterioration with or without unexplained weight loss, particularly in those with a personal or family history of autoimmune conditions (2). Disease progression is often measured by C-peptide testing, which functions as a marker of endogenous insulin production and can correlate with residual beta-cell function when collected appropriately with plasma glucose (2, 6).

Due to its autoimmune pathophysiology, the American Diabetes Association (ADA) aligns LADA treatment strategies closely with the management of type 1 diabetes, recommending early initiation of insulin for glycemic control (1). Despite this, disease variability in both phenotype and progression often warrants a multifaceted approach to pharmacologic treatment strategies, including both insulin and non-insulin antihyperglycemic options as indicated (2). A 2020 international consensus published by an expert panel reflected this more personalized approach, tailoring pharmacologic recommendations based on C-peptide levels at diagnosis, with a corresponding fasting glucose level of 80–180 mg/dL to demonstrate C-peptide appropriateness (2). Individuals with a C-peptide level less than 0.3 nmol/L are recommended to initiate a basal/bolus insulin regimen (2). Individuals with a C-peptide level between 0.3 and 0.7 nmol/L are recommended to initiate pharmacologic treatment more closely aligned with type 2 diabetes management, avoiding agents that may exhaust beta-cell function and expedite disease progression (2, 4). Individuals with a C-peptide level greater than 0.7 nmol/L are recommended to follow treatment guidelines for type 2 diabetes, although with repeat C-peptide measurements in the setting of worsening glycemic control (2).

Although this published consensus discusses the utility of non-insulin antihyperglycemic agents for the management of LADA, minimal data are currently available evaluating the use of these agents in individuals with LADA in clinical practice (2, 7). This study aimed to review antihyperglycemic prescribing patterns and relevant safety outcomes surrounding positive GAD-65 AAb detection at an academic integrated delivery network in Northeast Ohio.

Methods

Study design

A retrospective, observational cohort study was conducted at an academic integrated delivery network in Northeast Ohio using electronic medical records. Individuals were eligible for inclusion if they had a positive GAD-65 AAb test between September 1, 2021, and September 1, 2023, were at least 30 years of age at the time of GAD-65 AAb laboratory collection, and had a diagnosis code for type 2 diabetes mellitus in their electronic health record (EHR) before testing for GAD-65 AAb. Included participants were required to have had at least one primary care or endocrinology office visit or telehealth encounter within the health system during the two years before positive GAD-65 AAb laboratory collection and within 1 year after collection. Medication data were based on the medication list from the most recent encounter before GAD-65 AAb testing and the encounter closest to, but not after, 12 months after GAD-65 AAb testing. Medications could include both active prescriptions from the study institution and patient-reported medications from other health systems. The research protocol for this study was approved as Exempt Human Subject Research by the health system’s Institutional Review Board.

Most GAD-65 AAb values were precipitated by the reference laboratory, which used a GAD-65 laboratory test based on an enzyme-linked immunosorbent assay (ELISA) using recombinant human GAD-65 expressed in S. cerevisiae yeast. This was a bridge assay with a standard curve, with values reported in IU/mL, where the reference range was set at <5 IU/mL. Comparison of this GAD-65 ELISA with the radioimmunoassay (RIA) predicate demonstrated 97.1% positive agreement and 91.2% negative agreement.

Objectives

The primary objective of this study was to compare the median number of non-insulin antihyperglycemic medication classes prescribed immediately before positive GAD-65 AAb test, or at baseline, and at 12 months after detection. Secondary objectives were to describe the prescribed antihyperglycemic medication classes immediately before positive GAD-65 AAb test and at 12 months after detection. The frequency of C-peptide testing within 12 months after GAD-65 AAb detection was also described. Secondary outcomes included comparisons of the incidence of glycemic events, categorized as hypoglycemia and hyperglycemia, during the 12 months before and after GAD-65 AAb detection. Hypoglycemic events were defined as a blood glucose level of <54 mg/dL, an emergency department visit with a primary encounter diagnosis of hypoglycemia, or a hospital admission with a primary encounter diagnosis of hypoglycemia. Hyperglycemic events were defined as a blood glucose level of >300 mg/dL, an emergency department visit with a primary encounter diagnosis of hyperglycemia, or a hospital admission with a primary encounter diagnosis of hyperglycemia.

Statistical analysis

Descriptive statistics were used to summarize baseline characteristics. A univariate analysis using the Wilcoxon signed-rank test was performed to compare the median number of non-insulin antihyperglycemic medication classes prescribed immediately before positive GAD-65 AAb test and at 12 months after detection for the primary objective. Descriptive statistics were used for several secondary objectives, including the description of prescribed antihyperglycemic medication classes before and after positive GAD-65 AAb detection and the frequency of C-peptide monitoring at 12 months after GAD-65 AAb detection. A McNemar test was used to evaluate individual medication classes present before and after positive GAD-65 AAb detection and the frequency of C-peptide testing within 12 months after GAD-65 AAb detection. A univariate analysis using the McNemar test was also performed for secondary objectives, comparing the incidence of glycemic events during the 12 months before and after GAD-65 AAb detection.

Results

A total of 306 individuals had a positive GAD-65 AAb laboratory test collected between September 1, 2021, and September 1, 2023, across the index health system. The criterion for a positive GAD-65 AAb test was >5.0 IU/mL. Of these individuals, 195 met the inclusion criteria, with reasons for exclusion listed in Fig. 1.

Figure 1.

Study selection criteria. GAD, Glutamic Acid Decarboxylase; AAb, Autoantibody; T2DM, Type 2 Diabetes Mellitus.

Included participants had a mean age of 57 +/− 13.7 years (50.8% female, 73.3% white) and had a median BMI of 26.2 kg/m² at baseline. Most participants (82.1%) had a baseline eGFR of ≥60 mL/min/1.73 m². Additional baseline characteristics are included in Table 1. Notably, most participants did not have LADA mentioned in their chart before GAD-65 AAb detection, and 11.3% of participants had diagnosis codes for both type 2 diabetes and type 1 diabetes listed at the time of positive GAD-65 AAb detection.

Table 1.

Baseline characteristics (n = 195).

Age at the time of positive GAD-65 AAb, years, mean ± standard deviation	57.2 ± 13.7
Female sex, n (%)	99 (50.8)
Race, n (%)
- White	143 (73.3)
- Black	32 (16.4)
- Multiracial/multicultural	12 (6.2)
- Asian	2 (1.0)
- American Indian/Alaska Native	2 (1.0)
- Not listed	4 (2.1)
Hispanic or Latinx ethnicity, n (%)	22 (11.3)
BMI at the time of positive GAD-65 AAb, kg/m2, median (IQR)	26.17 (22.4–30.1)
BMI categorization, n (%)
- Underweight (<18.5 kg/m2)	5 (2.6)
- Healthy (18.5–24.9 kg/m2)	76 (39.0)
- Overweight (25–29.9 kg/m2)	63 (32.3)
- Class 1 obesity (30–34.9 kg/m2)	33 (16.9)
- Class 2 obesity (35–39.9 kg/m2)	6 (3.1)
- Class 3 obesity (>40 kg/m2)	10 (5.1)
- Not available	2 (1.0)
Past medical history, n (%)
- History of neuropathy	31 (15.9)
- History of retinopathy	25 (12.8)
- History of nephropathy	23 (11.8)
- History of chronic kidney disease	18 (9.2)
- History of stroke	15 (7.7)
- History of myocardial infarction	12 (6.2)
- History of heart failure	0 (0)
eGFR at the time of GAD-65 AAb, n (%)
- ≥60 mL/min/1.73 m2	160 (82.1)
- 45–59 mL/min/1.73 m2	10 (5.1)
- 30–44 mL/min/1.73 m2	12 (6.2)
- 15–29 mL/min/1.73 m2	4 (2.1)
- < 15 mL/min/1.73 m2	3 (1.5)
- Not available	6 (3.1)
LVEF at the time of GAD-65 AAb, %, mean ± standard deviation	56 ± 10.8
- Number of patients with LVEF in chart pre-GAD-65 AAb, n (%)	38 (19.5)
Patients with LADA in chart pre-GAD-65 AAb, n (%)	45 (23.1)
Patients with type 1 diabetes in chart pre-GAD-65 AAb, n (%)	22 (11.3)

GAD, glutamic acid decarboxylase; AAb, autoantibody; BMI, body mass index; IQR, interquartile range; eGFR, estimated glomerular filtration rate; LVEF, left ventricular ejection fraction; LADA, latent autoimmune diabetes in adults.

The median number of non-insulin antihyperglycemic agents prescribed was significantly lower 12 months after GAD-65 AAb detection than at the time of GAD-65 AAb laboratory collection. Prescriptions were identified from the medication list in the EHR at the designated time points. Figure 2 further illustrates the number of non-insulin antihyperglycemic medication classes at the time of GAD-65 AAb laboratory collection and 12 months after detection, along with the number of individuals who fell into each category. Figure 3 illustrates the same data as percentages of individuals in each category.

Figure 2.

Number of non-insulin antihyperglycemic classes pre- and post-GAD-65 AAb, as per number of participants (n = 195). GAD, Glutamic Acid Decarboxylase; AAb, Autoantibody.

Figure 3.

Number of non-insulin antihyperglycemic classes pre- and post-GAD-65 AAb, as per percentage of participants (n = 195). GAD, Glutamic Acid Decarboxylase; AAb, Autoantibody.

Table 2 shows the antihyperglycemic regimen at the time of positive GAD-65 AAb detection and 12 months after detection. When treatment patterns for prescribed antihyperglycemics at the time of GAD-65 AAb detection were compared with those 12 months after detection, fewer individuals had active prescriptions for biguanides (n = 104 vs n = 74, P < 0.0001), DPP-4 inhibitors (n = 16 vs n = 6, P = 0.0039), GLP-1 receptor agonists (n = 44 vs n = 34, P = 0.0499), SGLT-2 inhibitors (n = 36 vs n = 21, P = 0.0027), sulfonylureas (n = 43 vs n = 22, P = 0.0002), and thiazolidinediones (n = 6 vs n = 2, P = 0.0455) at 12 months after GAD-65 AAb detection compared with baseline. More individuals had a bolus insulin prescription at 12 months after GAD-65 AAb detection (n = 144) than at baseline (n = 108, P < 0.0001). More individuals had a basal insulin prescription at 12 months after GAD-65 AAb detection (n = 151) than at baseline (n = 146); however, this difference did not achieve statistical significance (P = 0.3173).

Table 2.

Antihyperglycemic regimen, pre- vs post-GAD-65 AAb.

	Pre-GAD-65 AAb	Post-GAD-65 AAb	P-value
Number of non-insulin antihyperglycemic classes, median (IQR)	1 (0–2)	1 (0–1.5)	P < 0.0001
Number of insulin classes, median (IQR)	2 (1–2)	2 (1–2)	P < 0.0001
Total number of antihyperglycemic classes, median (IQR)	2 (2–3)	2 (2–3)	P = 0.0005
Total patients on select antihyperglycemic class, n (%)
Alpha glucosidase inhibitor	0 (0)	0 (0)	P = 1.000
Amylin mimetic	0 (0)	0 (0)	P = 1.000
Biguanide	104 (53.3)	74 (37.9)	P < 0.0001
Bile acid sequestrant	0 (0)	1 (0.5)	P = 0.3173
DPP-4 inhibitor	16 (8.2)	6 (3.1)	P = 0.0039
Dopamine-2 agonist	0 (0)	0 (0)	P = 1.000
GLP-1 receptor agonist	44 (22.6)	34 (17.4)	P = 0.0499
GLP-1/GIP receptor agonist	4 (2.1)	7 (3.6)	P = 0.1797
Meglitinide	1 (0.5)	0 (0)	P = 0.3173
SGLT-2 inhibitor	36 (18.5)	21 (10.8)	P = 0.0027
Sulfonylurea	43 (22.1)	22 (11.3)	P = 0.0002
Thiazolidinedione	6 (3.1)	2 (1.0)	P = 0.0455
Basal insulin	146 (74.9)	151 (77.4)	P = 0.3173
Mixed insulin	5 (2.6)	3 (1.5)	P = 0.3173
Bolus insulin	108 (55.4)	144 (73.8)	P < 0.0001

GAD, glutamic acid decarboxylase; AAb, autoantibody; IQR, interquartile range; DPP-IV, dipeptidyl peptidase-4; GLP-1, glucagon-like peptide-1; GIP, glucose-dependent insulinotropic polypeptide; SGLT-2, sodium-glucose cotransporter 2.

The majority of individuals had a C-peptide value collected at the time of positive GAD-65 AAb detection or within the 12 months before (n = 166, 85.1%); however, significantly fewer individuals had a C-peptide measurement during the 12 months after GAD-65 AAb detection (n = 26, 13.3%, P < 0.0001). Of the C-peptide values collected, most were <0.3 nmol/L, both at baseline (n = 91, 54.8%) and during the 12 months after GAD-65 AAb detection (n = 12, 46.2%). Additional information about C-peptide data is presented in Table 3.

Table 3.

C-peptide monitoring.

	Pre-GAD-65 AAb	Post-GAD-65 AAb
Number of patients with a C-peptide drawn, n (%)	166 (85.1)	26 (13.3)
C-peptide value, n (%)*
- < 0.3 nmol/L	91 (54.8)	12 (46.2)
- 0.3–0.7 nmol/L	47 (28.3)	9 (34.6)
- > 0.7 nmol/L	28 (16.9)	5 (19.2)

GAD, glutamic acid decarboxylase; AAb, autoantibody.

^*Percentages of C-peptide testing stratified by C-peptide category were calculated among patients who underwent C-peptide testing at each time point, out of 166 patients pre-GAD-65 AAb and 26 patients post-GAD-65 AAb, respectively.

The incidence of glycemic events before and after GAD-65 AAb detection is outlined in Table 4 and stratified into hypoglycemic and hyperglycemic events according to the qualifying criteria. There was no statistically significant difference in hypoglycemic events during the 12 months before and after GAD-65 AAb detection. There was a significant decrease in hyperglycemic events from the 12 months before GAD-65 AAb detection (n = 94) to the 12 months after detection (n = 50, P < 0.0001).

Table 4.

Glycemic events, pre- vs post-GAD-65 AAb.

	Pre-GAD-65 AAb	Post-GAD-65 AAb	P-value
Patients who experienced hypoglycemic events, n (%)	15 (7.7)	19 (9.7)	P = 0.3173
Blood glucose < 54 mg/dL	13 (6.7)	19 (9.7)	P = 0.1336
ED visit for hypoglycemia	5 (2.6)	4 (2.1)	P = 0.6547
Hospital admission for hypoglycemia	2 (1.0)	2 (1.0)	P = 1.000
Patients who experienced hyperglycemic events, n (%)	94 (48.2)	50 (25.6)	P < 0.0001
Blood glucose > 300 mg/dL	93 (47.7)	48 (24.6)	P < 0.0001
ED visit for hyperglycemia	43 (22.1)	17 (8.7)	P < 0.0001
Hospital admission for hyperglycemia	34 (17.4)	11 (5.6)	P < 0.0001

GAD, glutamic acid decarboxylase; AAb, autoantibody; ED, emergency department.

Discussion

At 12 months after GAD-65 AAb detection, the median number of prescribed non-insulin antihyperglycemic agents was significantly lower than at the time of GAD-65 AAb laboratory collection. The median number of prescribed insulin agents was significantly higher, likely contributing to the observed reduction in hyperglycemic events during the 12 months after AAb detection. This clinically relevant secondary finding relates to the impact of GAD-65 AAb detection. Given the limited number of categorical groupings of prescribed antihyperglycemic classes, the median values appear the same before and after GAD-65 AAb detection; however, there were statistically significant reductions in the number of prescribed non-insulin antihyperglycemic agents following GAD-65 AAb detection. These findings suggest a trend toward reduced prescribing of non-insulin antihyperglycemic agents and increased prescribing of insulin after GAD-65 AAb detection; however, temporal association does not confirm causality, and additional research is needed to fully assess the existence and potential extent of these trends within the index health system.

Data regarding antihyperglycemic prescribing trends in clinical practice for individuals with type 2 diabetes and positive GAD-65 AAb are limited, restricting comparisons of these findings with prescribing patterns at other health systems. If a trend of reduced utilization of non-insulin antihyperglycemic agents after GAD-65 AAb detection is demonstrated in future studies, it would be important to investigate the appropriateness of deprescribing, particularly to determine whether non-insulin antihyperglycemics are being inappropriately discontinued in individuals with relevant comorbidities. Several antihyperglycemic agents are FDA approved for the management of disease states other than diabetes, including SGLT-2 inhibitors for chronic kidney disease (CKD) and congestive heart failure (CHF), as well as GLP-1 and GLP-1/GIP agonists for weight management and CKD (8, 9, 10, 11, 12, 13, 14, 15, 16, 17). For example, select SGLT-2 inhibitors and GLP-1 agonists have demonstrated a lower risk of eGFR decline and mortality from cardiorenal etiologies in individuals with CKD (8, 9, 14). Diabetes is a risk factor for the development and progression of CKD; therefore, individuals with positive GAD-65 AAb and CKD or risk factors for CKD may miss comorbid renal benefits if, for instance, select GLP-1 agonists are discontinued in favor of an insulin-based regimen upon autoantibody detection (14, 18). It is also recognized that some antihyperglycemic classes may be discontinued after risk–benefit analyses, such as discontinuation of SGLT-2 inhibitors due to the risk of euglycemic diabetic ketoacidosis (DKA). Therefore, further investigation into the appropriateness of antihyperglycemic prescribing trends with respect to comorbidities may support greater provision of individualized care in individuals with type 2 diabetes and positive GAD-65 AAb, particularly as it pertains to expansion of shared-decision-making and integration of multidisciplinary care.

Incidence of baseline C-peptide testing was appropriately collected in most participants, with appropriate collection being deemed in tandem with a plasma glucose value (19). However, incidence of repeat C-peptide testing within 12 months after GAD-65 AAb detection was low, occurring in only 13.3% of participants, illuminating deviation from expert recommendations to repeat C-peptide measurements at 6 months (1). Adherence to repeat C-peptide testing is imperative for tailoring individualized glycemic regimens, as individuals with C-peptide values >0.3 nmol/L may have delayed requirements for exogenous insulin. Thus, the low incidence of repeat C-peptide testing is a clinically relevant finding as it pertains to factors influencing pharmacotherapy decision-making and resultant glycemic outcomes. This introduces an opportunity to leverage non-insulin antihyperglycemics that may afford multimorbidity benefits, particularly among those with obesity, CKD, cardiovascular disease (CVD), and CHF, where many of these therapies reduce the risk of adverse outcomes independent of their effects on glycemic control. A statistically significant decline in sulfonylurea prescribing from pre- to post-GAD-65 AAb detection was found in this study, aligning with expert guidance to avoid therapies that exhaust beta-cell function (2). As beta-cell function declines, repeat C-peptide testing becomes imperative to identify when exogenous insulin may be required to prevent hyperglycemic crisis, such as diabetic ketoacidosis (2). Further investigation is warranted to identify any barriers to repeat C-peptide testing in practice to optimize adherence to expert recommendations and potentially better tailor a patient-specific, C-peptide-stratified approach when constructing antihyperglycemic therapeutic plans.

The inclusion criteria were a study strength, where selection of 30 years or greater as the age cutoff aligned with LADA diagnostic criteria, thus increasing the likelihood of capturing individuals with potential LADA versus other autoimmune diabetes in younger populations (2). Patient eligibility was based on positive GAD-65 AAb and a historic type 2 diabetes diagnosis rather than solely including patients based on the presence of LADA diagnosis via ICD codes. If electing to do the latter, numerous eligible patients would have likely been excluded, due to the observed lower frequency of LADA entry into EHR problem lists. Furthermore, several individuals had conflicting diabetes diagnosis codes listed in their chart at the time of positive GAD-65 AAb laboratory collection, with 23.1% having both LADA and type 2 diabetes listed and 11.3% having both type 1 and type 2 diabetes listed. This proportion of individuals with conflicting diagnosis codes highlights the inaccuracy and unreliability of diagnosis code entry to reflect true diagnosis, further supporting our rationale for GAD-65 AAb detection as a qualifying component for inclusion criteria.

Included participants were required to have a previous diagnosis of type 2 diabetes to minimize the risk of capturing patients with type 1 diabetes instead, highlighting their slower progression of dysglycemia as seen with LADA versus a more rapid decline as seen with type 1 diabetes (2). Utility of positive GAD-65 AAb testing versus LADA diagnosis entry also increases the likelihood that antihyperglycemic regimen changes reflect real-time prescribing responses to the positive GAD-65 AAb, whereas time of LADA diagnosis entry into the EHR may not reflect true time of diagnosis and thus changes to antihyperglycemic regimens may not truly reflect therapeutic changes due to the diagnosis.

Timing of post-positive GAD-65 AAb antihyperglycemic regimen assessment at 12 months after testing was also a study design strength, increasing the likelihood of capturing visits where the positive GAD-65 AAb result was addressed, enabling inclusion of patients who have less frequent appointments. Requiring primary care or endocrinology visits both prior to and after GAD-65 AAb detection was also an inclusion criterion strength, increasing the likelihood that included individuals were actively following with the health system and thus changes to prescribed antihyperglycemic regimens due to GAD-65 AAb detection were more likely to be captured. The retrospective nature of this study limits the internal validity of its findings, generating an inability to account for several confounding variables due to limitations in data collection and hindrance of seeking justifications via chart review. The retrospective nature of the study limited uniform collection of variables that may otherwise provide clinically meaningful contributions to the subject matter, including omission of glycemic data collection and endocrinology referral data, both of which may be of benefit to include in future study designs. This also hindered our ability to collect initial clinical indication for GAD-65 AAb test ordering, as the retrospective study design meant provider rationale for laboratory order placement could not be uniformly extracted from the EHR. Our study did not explicitly account for antihyperglycemics present solely for comorbid indications; although select comorbidities were described, their proportionate impact on antihyperglycemic prescribing was not well-articulated in the context of the primary outcome and select secondary outcomes. Our study operated under the assumption that changes in antihyperglycemic prescribing were a direct result of positive GAD-65 AAb results; thus, other variables influencing therapeutic changes were not accounted for, highlighting an important limitation. For instance, the decrease in non-insulin antihyperglycemic quantity and increase in insulins post-GAD-65 AAb may be driven by progressive β-cell failure and worsening dysglycemia, likely prompting GAD-65 AAb testing, rather than directly due to the positive AAb test result itself. Detection of other autoimmune markers of diabetes was not assessed, where the quantity and composition of autoimmunity may have influenced treatment decisions beyond our scope of study detection. The utility of GAD-65 AAb testing may have also overlooked patients with advanced autoimmunity, where the presence of autoantibodies can diminish with pathogenic progression, thus potentially underestimating our study population and not representing those with severely progressed pathogenesis (2). Conversely, selection of GAD-65 AAb detection may also overestimate the true incidence of LADA vs adult-onset type 1 diabetes in our study population, as this study did not factor in other diagnostic criteria for LADA, such as the absence of insulin requirements for 6 months post-GAD-65 AAb detection (2). The potential heterogeneity of the underlying diabetes pathophysiology, a study cohort that is predominantly white, and a single-center study design limit the generalizability of study findings to more diverse populations. Limitations are present surrounding our secondary outcome of glycemic event analysis. The true incidence of hypoglycemia and hyperglycemia is likely underrepresented, as instances of glycemic excursions may be underdocumented and glycemic events occurring at home or outside health systems would not have been captured.

Conclusion

In patients previously diagnosed with type 2 diabetes, GAD-65 AAb detection was associated with a shift toward insulin-based management, characterized by a decrease in the median number of non-insulin antihyperglycemics prescribed and an increase in the median number of insulin agents prescribed at 12 months after GAD-65 AAb detection. Further research is warranted to confirm any definitive differences in treatment patterns before and after GAD-65 AAb detection, as well as the appropriateness of any observed prescribing patterns as they relate to glycemic control and relevant comorbidities.

Declaration of interest

Kevin M Pantalone, DO, ECNU, FACE, received consulting fees from Bayer, Boehringer Ingelheim, Corcept Therapeutics, Diasome, Eli Lilly, Merck, Novo Nordisk, and Sanofi; receives speakers’ honoraria from AstraZeneca, Corcept Therapeutics, and Novo Nordisk; receives research support from Bayer, Novo Nordisk, and Twin Health; and has a patent application, ‘Identifying Patients for Intensive Hyperglycemia Management’ (US Provisional Patent Application No. 62/982,195). All other authors have no known conflicts of interest to disclose, financial or otherwise, affiliated with this research report or subsequent manuscript.

Funding

This research did not receive any specific grant from any funding agency in the public, commercial, or not-for-profit sector.

Ethical approval (human study)

The study was approved by the Cleveland Clinic Institutional Review Board.