Risk of clinically relevant hyperglycemia with metoprolol compared to carvedilol in older adults with heart failure and diabetes

Abstract

Background:

Although prior literature suggests that metoprolol may worsen glucose control compared to carvedilol, whether this has clinical relevance among older adults with diabetes and heart failure remains an open question.

Methods:

This was a US retrospective cohort study utilizing data sourced from a 50% national sample of Medicare fee-for-service claims of patients with Part D prescription drug coverage (2007–2017). Among patients with diabetes and heart failure we identified initiators of metoprolol or carvedilol, which were 1:1 propensity score matched on >90 variables. The primary outcome was initiation of a new oral or injectable antidiabetic medication (proxy for uncontrolled diabetes); secondary outcomes included initiation of insulin and severe hyperglycemic event (composite of emergency room visits or hospitalizations related to hyperglycemia).

Results:

Among 24,239 propensity score-matched pairs (mean [SD] age 77.7 [8.0] year; male (39.1%)), there were 8,150 [incidence rate per 100 person-years (IR) = 33.5] episodes of antidiabetic medication initiation among metoprolol users (exposure arm) compared to 8,576 [IR = 33.4] among carvedilol users (comparator arm) compared to corresponding to an adjusted hazard ratio [aHR] of 0.97 (95% CI: 0.94, 1.01). Similarly, metoprolol was not associated with a significant increase in the risk of secondary outcomes including insulin initiation: aHR of 0.98 (95% CI: 0.93, 1.04)) and severe hyperglycemic events: aHR of 0.98 (95% CI: 0.93, 1.02).

Conclusions:

In this large study of older adults with heart failure and diabetes, initiation of metoprolol compared to carvedilol was not associated with an increase in the risk of clinically relevant hyperglycemia.

The prevalence of diabetes among older adults is estimated to be greater than 25%, the highest of any age group.¹ Older adults diagnosed with diabetes have higher rates of mortality, reduced functional status, increased risk of institutionalization,² and are more likely to develop diabetic complications.^3,4 The goal of diabetes management – to prevent acute and chronic diabetic complications while minimizing harms – is substantially complicated among older adults due to the high prevalence of multimorbidity,⁵ polypharmacy,⁶ geriatric syndromes such as frailty and sarcopenia,⁷ functional and cognitive impairment,⁶ and age-specific physiological changes related to ageing that can compromise the adaptive responses to high serum glucose, predisposing them to chronic hyperglycemia. ^8–11

With an estimated prevalence ranging between 22% to 27%, heart failure (HF) is common among older adults with diabetes.^12–14 The putative mechanism for this high prevalence is multifactorial and often attributed to the shared pathophysiological factors underpinning the development of both conditions including chronic hypertension, coronary artery disease, renal insufficiency, and obesity.¹⁵ Despite initial concerns that β-blockers may worsen glucose control and therefore be inappropriate in patients with diabetes,¹⁶ they have consistently demonstrated a survival benefit in diabetic patients with HF with reduced ejection fraction (HFrEF)¹⁷ and have become the cornerstone of pharmaceutical management in HFrEF patients regardless of diabetes status. However, not all agents within the class are thought to worsen glucose control to a similar extent. Currently, metoprolol and carvedilol are the two most commonly used agents in the United States for HFrEF, with current guidelines showing no clear preference for one over the other.^18,19 Findings from animal studies suggest that metoprolol (compared to carvedilol) may worsen glucose control in patients with diabetes. Metoprolol may induce acute hyperglycemia by reducing insulin sensitivity via an acute deterioration of insulin-stimulated endothelial function.²⁰ Over the long-term, metoprolol may cause greater weight-gain contributing to insulin resistance.^21,22 Conversely, carvedilol is thought to have more favorable effects on glucose metabolism via inhibition of the α-adrenergic receptor.²²

Whether this contrast in pharmacological properties translates to clinically relevant differences in glucose control in older adults receiving evidenced based care for HF and diabetes remains unclear. Accordingly, we aimed to assess whether initiation of metoprolol compared to carvedilol is associated with an increase in the risk of clinically relevant hyperglycemia in older adults.

METHODS

Data Sources and Study Population

Study subjects were drawn from Medicare insurance claims, a US federal program which provides healthcare to US citizens over 65 years of age. More specifically, a 50% sample of Medicare fee-for-service data with linked Part D prescription claims from January 2007 to December 2017 were utilized. Data elements used included patient sociodemographic characteristics, medical and pharmacy enrollment status, inpatient and outpatient medical services (International Classification of Disease, Ninth and Tenth Revision; Current Procedural Terminology, Fourth Edition), and outpatient pharmacy dispensing (drug name and strength, units dispensed, and days’ supply).

We identified a cohort of patients initiating metoprolol or carvedilol; the date of initial medication dispensing was designated as the index date. Cohort membership required patients to be new users of metoprolol or carvedilol, without any prior β-blocker therapy in the baseline period (defined as 180-day period prior to and including the date of cohort entry), be ≥ 65 years of age at cohort entry, and have used oral or injectable antidiabetic medications during the baseline period. We required patients to have a hospitalization with an inpatient diagnosis of HF during the baseline period at any position; in prior validation studies, the positive predictive value for this approach exceeds 80%.²³ Patients with diagnosis of cancer or human immunodeficiency virus, evidence of dysphagia or nasogastric intubation, or those initiating both metoprolol and carvedilol on the same day were excluded from analysis.

The study was approved by the Rutgers University Institutional Review Board, and the appropriate data use agreements were in place.

Propensity Score Matching

All analyses were performed using SAS 9.4 (SAS Institute Inc, Cary, NC). To mitigate the risk of confounding, new initiators of metoprolol (exposure) were matched to carvedilol (control) initiators on their estimated propensity score which modelled the probability of initiating metoprolol using >90 baseline covariates assessed during the baseline period (please see Appendix Table 1 for list of variables and their definitions). Propensity scores were comprised of variables corresponding to the domains of sociodemographics (e.g., age, sex, race), complications of diabetes (e.g., diabetic neuropathy, retinopathy, nephropathy), oral and injectable glucose lowering therapy use (e.g., metformin, insulin, dipeptidyl peptidase-4 inhibitors), cardiovascular conditions (e.g., myocardial infarction, stroke, heart failure), cardiovascular medication use (e.g., dispensing of angiotensin-converting enzyme inhibitors, calcium channel blockers, thiazides and loop diuretics, statins), non-cardiovascular comorbid conditions (e.g., diagnosis of chronic kidney disease, chronic obstructive pulmonary disease, asthma, obstructive sleep apnea, anxiety, depression), and non-cardiovascular medication use (e.g., dispensing of anticonvulsants, antidepressants, antianxiety, oral steroids, antipsychotics). A 1:1 propensity score-matched cohort was created utilizing the nearest neighbor matching approach with a maximum caliper width of 0.01.²⁴

Follow Up and Study Endpoint

Patients began contributing follow-up time from the day after cohort entry up until the first occurrence of the following: end of healthcare or pharmacy eligibility, switching to the comparator (i.e., patients in the metoprolol arm switching to carvedilol and vice versa), therapy discontinuation (defined as a 30-day gap in treatment after the last prescription of the study drugs), end of study data (December 31, 2017), or the occurrence of the outcome.

The primary study endpoint was clinically relevant hyperglycemia, operationalized as the initiation of a new oral or injectable antidiabetic medication therapy from one of the following drug-classes: metformin, sulfonylureas, dipeptidyl peptidase 4 inhibitors, glucagon-like peptide-1 receptor agonists, sodium-glucose co-transporter-2 inhibitors, thiazolidinediones, alpha glucosidase inhibitors, meglitinide derivatives, or insulin. In order to qualify as new therapy, patients could not have used any agents from the same medication class during the baseline period. Instances of switching from one agent to another within the same class was not considered as initiation of new antidiabetic therapy.

Primary Analysis

We assessed the performance of the propensity score by cross-tabulating the covariates prior to – and after – propensity score matching by exposure group and utilizing a threshold of 10% in standardized difference as a meaningful imbalance between the two groups.²⁵ We estimated the risk of antidiabetic medication initiation for metoprolol compared to carvedilol by calculating the number of events, incidence rates per 100 person-years, and cox proportional hazard models were used to generate hazard ratios (HRs) with the Wald 95% confidence intervals (CI). Kaplan-Meier curves were generated to visualize the risk of the outcome over time and log-rank tests were used to compare the survival distribution in the two groups.

Sensitivity Analyses and Secondary Outcomes

To assess the robustness of the primary findings, we conducted sensitivity analysis pertaining to exposure-related censoring criteria, where instead of censoring patients at the time of treatment switching or discontinuation, we instead carried the index treatment forward to mimic an intention-to-treat approach. Furthermore, we employed a broader HF definition also allowing patients who had two outpatient diagnosis of HF during the baseline period to contribute to the analysis.²⁶

Three secondary outcomes were examined. First, because the initiation of insulin represents a significant escalation of therapy in diabetes, we examined insulin initiation as an outcome. For this outcome, analysis was restricted to patients who were insulin naïve during the baseline period. Second, among the subgroup of patients on metformin monotherapy, we assessed the risk of initiating other second-line antidiabetic therapies. Finally, we examined risk of a severe hyperglycemic event defined as a composite of emergency department visits or hospitalizations related to hyperglycemia (including codes relating to diabetic ketoacidosis and hyperosmolar hyperglycemic state; see Appendix Table 1 for list of codes).^9,27

Finally, we tested for the presence of effect modification in subgroups of patients including by age, gender, and severity of heart failure. In instances when the cohort was recreated (i.e., for the outcome of insulin initiation, initiation of second-line antidiabetic therapy in patients using metformin monotherapy and the use of the broader HF criterion or subgroups), propensity scores were re-estimated, and patients were 1:1 matched on their newly estimated propensity score using the nearest neighbor approach as described above.

Role of the Funding Source

The study was funded by the New Jersey Alliance for Clinical and Translational Science. The sponsor had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

RESULTS

Within the data, there were 3,335,253 initiators of metoprolol or carvedilol, out of which 608,036 also had evidence of antidiabetic medication use. After requirement of heart failure hospitalization was applied, there were 99,787 remaining patients. The application of the exclusion criteria (HIV or cancer) resulted in the final sample size of 61,675 patients that were eligible for propensity score matching.

Table 1 shows the distribution of select baseline characteristics prior to and after propensity score matching (see Appendix Table 2 for information on all baseline variables). After application of inclusion and exclusion criteria, there were 35,801 patients initiating metoprolol compared to 25,874 patients initiating carvedilol. With a few exceptions, most baseline characteristics were well balanced between the two groups (defined as standardized difference of less than 10%) even prior to propensity score matching. Patients initiating carvedilol were slightly younger, more likely to be male, and less likely diagnosed with atrial fibrillation. 1:1 propensity score matching on >90 baseline covariates resulted in a match of nearly 96% of patients in the carvedilol group (the smaller of the two groups) to a patient in the metoprolol group for a total of 25,341 matched pairs. After matching, baseline characteristics were closely balanced between the two groups with no standardized difference exceeding 10%.

Table 1:

Select baseline characteristics prior to and after propensity score matching ¹

	Prior to matching			After matching
	Metoprolol (n=35,801)	Carvedilol (n=25,874)	SD	Metoprolol (n=25,341)	Carvedilol (n=25,341)	SD
	Sociodemographic, n (%)
Age, mean (St.D)	78.3 (8.2)	77.4 (8.0)	11.1	77.7 (8.0)	77.7 (8.0)	0.0
Male	12,429 (34.7)	10,532 (40.7)	12.4	9,473 (39.1)	9,483 (39.1)	0.1
White	26,640 (74.4)	18,123 (70.0)	9.8	17,264 (71.2)	17,211 (71.0)	0.5
Black	4,702 (13.1)	4,102 (15.9)	7.7	3,667 (15.1)	3,680 (15.2)	0.1
Other Race	4,459 (12.5)	3,649 (14.1)	4.9	3,308 (13.6)	3,348 (13.8)	0.5
Diabetic complications and medications, n (%)
Diabetic neuropathy	10,013 (28.0)	6,993 (27.0)	2.1	6,599 (27.2)	6,572 (27.1)	0.3
Diabetic retinopathy	3,419 (9.6)	2,686 (10.4)	2.8	2,450 (10.1)	2,455 (10.1)	0.1
Diabetic nephropathy	6,390 (17.8)	4,852 (18.8)	2.3	4,488 (18.5)	4,469 (18.4)	0.2
Insulin	17,970 (50.2)	13,098 (50.6)	0.9	12,263 (50.6)	12,345 (50.9)	0.7
Metformin	14,754 (41.2)	10,416 (40.3)	1.9	9,806 (40.5)	9,764 (40.3)	0.4
DPP-4 inhibitors	3,970 (11.1)	2,941 (11.4)	0.9	2,725 (11.2)	2,760 (11.4)	0.5
Cardiovascular disease and medication use, n (%)
Stroke	9,940 (27.8)	6,487 (25.1)	6.1	6,238 (25.7)	6,243 (25.8)	0.0
Myocardial Infarction	10,152 (28.4)	7,431 (28.7)	0.8	6,976 (28.8)	7,039 (29.0)	0.6
Unstable angina	5,750 (16.1)	3,939 (15.2)	2.3	3,751 (15.5)	3,772 (15.6)	0.2
Other Ischemic Heart Disease	24,227 (67.7)	18,746 (72.5)	10.5	17,295 (71.4)	17,355 (71.6)	0.5
Peripheral vascular disease	11,167 (31.2)	7,790 (30.1)	2.4	7,372 (30.4)	7,354 (30.3)	0.2
Atrial Fibrillation	17,939 (50.1)	10,623 (41.1)	18.2	10,373 (42.8)	10,393 (42.9)	0.2
Other Dysrhythmias	16,709 (46.7)	11,173 (43.2)	7.0	10,625 (43.8)	10,608 (43.8)	0.1
ACE inhibitors	13,221 (36.9)	9,879 (38.2)	2.6	9,162 (37.8)	9,152 (37.8)	0.1
ARB II blockers	5,993 (16.7)	4,319 (16.7)	0.1	4,049 (16.7)	4,067 (16.8)	0.2
Calcium channel blockers	8,840 (24.7)	6,278 (24.3)	1.0	5,929 (24.5)	5,906 (24.4)	0.2
Non-dihydropyridine CCB	5,507 (15.4)	3,313 (12.8)	7.4	3,228 (13.3)	3,163 (13.0)	0.8
Thiazide diuretics	6,975 (19.5)	4,940 (19.1)	1.0	4,655 (19.2)	4,673 (19.3)	0.2
Loop diuretics	19,221 (53.7)	14,200 (54.9)	2.4	13,216 (54.5)	13,127 (54.2)	0.7
Statins	18,147 (50.7)	12,755 (49.3)	2.8	12,000 (49.5)	11,985 (49.4)	0.1
Comorbid conditions and medication use, n (%)
Asthma	6,073 (17.0)	3,824 (14.8)	6.0	3,687 (15.2)	3,701 (15.3)	0.2
COPD	17,197 (48.0)	11,833 (45.7)	4.6	11,247 (46.4)	11,162 (46.0)	0.7
Obstructive sleep apnea	5,196 (14.5)	3,444 (13.3)	3.5	3,292 (13.6)	3,289 (13.6)	0.0
Depression	10,475 (29.3)	6,549 (25.3)	8.9	6,332 (26.1)	6,368 (26.3)	0.3
Anxiety	6,525 (18.2)	3,949 (15.3)	7.9	3,824 (15.8)	3,892 (16.1)	0.8
Oral steroids	7,082 (19.8)	4,584 (17.7)	5.3	4,400 (18.2)	4,402 (18.2)	0.0
Antidepressants	11,853 (33.1)	8,041 (31.1)	4.4	7,634 (31.5)	7,732 (31.9)	0.9
Antipsychotics	3,891 (10.9)	2,354 (9.1)	5.9	2,296 (9.5)	2,325 (9.6)	0.4

Abbreviations: ACE: Angiotensin-converting enzyme; ARB: Angiotensin II receptor blocker; CCB: Calcium Channel Blockers; COPD: Chronic obstructive pulmonary disease; DPP-4: Dipeptidyl Peptidase-4; SD: Standardized Differences as percentages; St.D: Standard Deviation;

¹The table presents select pooled baseline characteristics before and after propensity score matching. See Appendix Tables 2 for information on all baseline covariates prior to and after propensity score matching.

Primary Analysis

Table 2 shows the number of events, incidence rates, and HRs for the primary analysis prior to and after matching. Prior to matching and over approximately 13 months of mean follow up, there were 12,060 episodes of antidiabetic medication initiation (incidence rate per 100 person-years [IR] = 33.7) among metoprolol users compared to 9,205 initiations [IR = 33.3 among carvedilol users corresponding to an unadjusted HR of 0.97 (95% CI: 0.95, 1.00). After matching, there were 8,150 [IR = 33.5] events in the metoprolol group compared to 8,576 [IR = 33.4] events in the carvedilol group, corresponding to an adjusted HR of 0.97 (95% CI: 0.94, 1.01). Prior to matching, the average daily dose of metoprolol and carvedilol over follow up were 62.9 mg and 20.3 mg, respectively and were 62.5 mg and 20.2 mg, respectively after matching. Figure 1 shows the pooled cumulative incidence of the primary outcome in the matched cohort over time along with the p-value for the log-rank test showing no clinically meaningful difference between the two groups at any period of follow up (p=0.10).

Table 2:

Risk of initiation of new antidiabetic therapies with metoprolol versus carvedilol prior to and after propensity score matching

	Prior to matching		After matching 1
	Metoprolol (n=35,801)	Carvedilol (n=25,874)	Metoprolol (n=25,341)	Carvedilol (n=25,341)
Events (IR)	12,060 (33.7)	9,205 (33.3)	8,150 (33.5)	8,576 (33.4)
Mean (SD) follow-up, months 2	12.5 (1.7)	13.6 (1.8)	12.8 (1.7)	13.5 (1.8)
Hazard Ratio (95% CI)	0.97 (0.95, 1.00)		0.97 (0.94, 1.01)

Abbreviations: CI: Confidence Interval; IR: Incidence Rates per 100 person-years of follow up; SD: Standard deviation

¹Propensity score assessed and adjusted for more than 90 covariates at baseline. Carvedilol is the reference group. See Appendix Table 2 for more information.

²Standardized differences between the two groups were less than 10% prior to and after propensity score matching.

Figure 1:

Propensity score-matched Kaplan-Meier curves for cumulative incidence of initiation of new antidiabetic therapy with metoprolol compared to carvedilol.

Sensitivity Analyses and Secondary Outcomes

Table 3 shows the findings from the sensitivity and secondary analyses. Changing the analysis from an as-treated to intention-to-treat approach did not change the point estimates appreciably. The risk of the primary outcome was similar for all subgroups and in the cohort where the broader HF definition was employed. Findings were similar for secondary outcomes; metoprolol was not associated with a significant increase in risk of insulin initiation, initiation of second-line antidiabetic therapies in patients using metformin monotherapy or severe hyperglycemic events.

Table 3:

Adjusted sensitivity and secondary analyses ¹

	No. of patients 2	Metoprolol Events (IR)	Carvedilol Events (IR)	HR (95% CI)
Primary Analysis	24,239	8,150 (33.5)	8,576 (33.4)	0.97 (0.94, 1.01)
Subgroup analysis 3
Male	9,408	3,279 (37.2)	3,472 (36.1)	0.98 (0.94, 1.03)
Female	14,739	4,965 (32.1)	5,077 (31.8)	0.99 (0.95, 1.03)
Age > 80 years	8,901	2,945 (36.6)	3,051 (37.9)	0.96 (0.91, 1.01)
Digoxin or spironolactone use	4,286	1,426 (33.5)	1,415 (31.8)	1.03 (0.96, 1.11)
Recent hospitalization (30 days)	14,436	4,654 (32.1)	5,022 (32.8)	0.95 (0.91, 0.99)
Sensitivity analysis
Intention to treat analysis	24,239	9,683 (26.9)	9,716 (28.1)	0.98 (0.95, 1.01)
Broad HF definition 3	56,060	19,878 (31.6)	20,920 (31.7)	0.97 (0.95, 0.99)
Secondary Outcomes
Insulin initiation 3,4	11,840	2,676 (16.3)	2,845 (16.3)	0.98 (0.93, 1.03)
Second-line therapy initiation 3,5	3,754	1,549 (41.2)	1,641 (44.4)	0.94 (0.87, 1.00)
Hyperglycemic hospitalization	24,239	3,396 (9.8)	3,658 (9.8)	0.99 (0.94, 1.03)
Emergency room visit	24,239	868 (2.3)	962 (2.3)	0.97 (0.88, 1.06)
Severe hyperglycemic event 6	24,239	3,822 (11.2)	4,151 (11.3)	0.98 (0.93, 1.02)

Abbreviations: CI: Confidence Interval; IR: Incidence Rates per 100 person-years of follow up; HF Heart Failure

¹Analyses were adjusted using propensity score matching. See text for details.

²In each group (represents the number of matched pairs).

³Propensity score was re-estimated and patients were re-matched

⁴Comprised of insulin naïve patients

⁵Initiation of second-line antidiabetic therapy was assessed in patients who were on metformin monotherapy at baseline.

⁶Composite of inpatient or emergency department visits relating to hyperglycemia; see Appendix Table 1 for codes.

DISCUSSION

It has been previously postulated that based on its mechanistic properties, metoprolol increases the risk of clinically relevant hyperglycemia in patients with HF and diabetes compared to carvedilol; however, whether this postulate is of clinical importance is unclear and to our knowledge this association has not been studied in a routine care setting. This study examined a Medicare-insured cohort of older adults with HF and diabetes initiating either metoprolol or carvedilol; the study found that metoprolol was not associated with an increase in the risk of new antidiabetic medication initiation, an indicator of clinically relevant hyperglycemia. Findings were consistent across multiple sensitivity analyses and for the secondary outcomes of insulin initiation and severe hyperglycemia.

The study has important clinical implications. Because metabolic impairment is intrinsic to HF pathophysiology, glucose control is harder to achieve in patients with HF and diabetes;^28,29 therefore, the identification and avoidance of agents that further increase serum glucose are relevant in guiding patient care. In addition, agents that induce hyperglycemia may inadvertently lead to a prescribing cascade where additional glucose-lowering therapies are added to mitigate the unintended adverse reaction (e.g., hyperglycemia) of a precipitating agent rather than switch to a more appropriate one³⁰; such practices further contribute to the increasing prevalence of polypharmacy in older adults, who are especially predisposed to developing drug-related adverse reactions. Our findings suggest that metoprolol does not meaningfully worsen hyperglycemia or contribute to such prescribing cascades in patients with HF and diabetes.

Although data on the individual β-blockers and risk of hyperglycemia in the setting of co-occurring HF and diabetes are lacking, prior studies have examined the hyperglycemic potential of metoprolol compared to carvedilol in other populations. Our findings are similar to the GEMINI trial which randomized patients with a diagnosis of hypertension and diabetes to carvedilol (n = 498) and metoprolol (n = 737) and found that over 8-months of follow up, patients initiating carvedilol had a statistically significant smaller magnitude of increase in hemoglobin A1c compared to those initiating metoprolol (0.02% v 0.15% corresponding to a modest difference of 0.13% in A1c between the two groups); however, this difference was not clinically meaningful.²² The mean age of the participants in the trial was 61 years compared to 78 years in our study, and the prevalence of HF was not reported. Notably, the study did not assess for differences in the rate of antidiabetic medication initiation or other clinically relevant endpoints relating to hyperglycemia. By contrast, our findings differ from a recently published observational study of older US nursing home patients initiating “diabetes-friendly” (carvedilol, nebivolol and labetalol) versus “diabetes-unfriendly” (atenolol, bisoprolol and metoprolol) β-blockers (n = 765 1:1 propensity score matched pairs) in patients with a diagnosis of myocardial infarction and diabetes. The study found a significantly lower rate of hyperglycemic hospitalizations, odds ratio 0.45 (95% CI: 0.21–0.97) with carvedilol use compared to this study which found no association with hyperglycemia-related hospitalizations.²⁷ This discrepancy in findings may be due to differences in source population (i.e., frail older institutionalized patients with myocardial infarction compared to older community dwelling adults with heart failure) or due to the differences in the beta-blockers that were studied. Our study focused on the two most frequently used evidence-based beta-blockers (metoprolol and carvedilol), unlike the prior study, which also included other lesser-used agents such as labetalol and nebivolol that were grouped based on their hyperglycemic potential.

Study findings should be viewed in light of limitations. First, because this is an observational study, it is susceptible to residual confounding. For instance, although we controlled for over 90 pertinent variables that could be potentially associated with both choice of β-blocker therapy and risk of hyperglycemia, we did not have access to some important variables related to diabetes (e.g., body mass index or hemoglobin A1c), HF (e.g., New York heart association functional classification or ejection fraction) or socioeconomic status (e.g. income levels). However, at least for diabetes-related variables, a prior study that linked Medicare data to electronic health records showed that claims-based proxies adequately adjusted for these unmeasured characteristics.³¹ Furthermore, we noted that several key variables were well balanced in the metoprolol and carvedilol groups even prior to propensity score matching, and we were able to match >95% of patients (in the smaller of the two arms), indicating good therapeutic equipoise and interchangeable use in routine clinical care, mitigating – but not eliminating – the concern for residual confounding. Second, study findings are most generalizable to older adults enrolled in Medicare fee-for-service plans, although we would not expect the biological effects of metoprolol to vary by insurance status. Third, because β-blocker therapy is often initiated at a lower dose and titrated to a target (or maximum tolerated) dose, we did not examine a dose-response relationship. Fourth, we did not have access to ejection fraction and therefore did not stratify analysis by HF subtype (reduced or preserved ejection fraction). However, β-blocker therapy is not routinely recommended for HF with preserved ejection fraction (HFpEF) in the absence of an alternative indication (e.g., chronic coronary syndromes).³² Furthermore, it is unlikely that beta-blocker induced glucose-alterations would vary significantly between patients with HFrEF and HFpEF. Fifth, the study utilized the initiation of antidiabetic therapies as a proxy for clinically relevant hyperglycemia, meaning that patients switching to other similarly efficacious glucose lowering therapies due to tolerability concerns would be misclassified as being uncontrolled. However, because we show similar findings for the secondary outcomes of insulin initiation (and second-line therapy initiation for metformin monotherapy) and for the more serious outcome of severe hyperglycemia, we do expect this to substantially influence our study findings. Finally, prior validation studies of HF have primarily utilized hospitalizations-based algorithms to validate HF (i.e. low sensitivity), meaning that it is possible that study findings may generalize more to the subset of patients with a more severe HF symptomology. However, prior studies have demonstrated that hospitalizations are common for patients with HF, ³³ and a sensitivity analysis including patients with HF diagnosed in the outpatient setting found similar results assuaging such concerns.

Overall, this study offers an important step in understanding the role of the two most commonly utilized, evidence-based β-blockers, carvedilol and metoprolol, on the risk of developing clinically relevant hyperglycemia in older adults with HF and diabetes. The study provides reassuring data for the use of metoprolol in these patients and suggests that factors other than glucose control should be considered when selecting an evidence-based β-blocker for HF in patients with diabetes

Key points.

• Metoprolol and carvedilol are the two most commonly used β-blockers in the US.

• Pharmacodynamic data suggests that metoprolol – compared to carvedilol – may worsen glucose control; however, the clinical relevance of these findings in older adults with pre-existing diabetes mellitus (DM) and heart failure (HF) is unclear.

• A 50% random sample of Medicare data were used to compare the risk of clinically relevant hyperglycemia among older adults ≥65 with evidence of DM and HF initiating these agents.

• The study found no increase in the risk of hyperglycemia in patients using metoprolol compared to carvedilol.