Differences in Complications, Cardiovascular Risk Factor, and Diabetes Management among participants enrolled at Veterans Affairs (VA) and Non-VA Medical Centers in the Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE)

Abstract

Aims:

We evaluated differences in participants with type 2 diabetes (T2DM) enrolled in the GRADE study at VA vs non-VA sites, focusing on cardiovascular risk factors and rates of diabetes care target achievements.

Methods:

We compared baseline characteristics between participants at VA (n=1216) and non-VA (n=3831) sites, stratifying analyses by cardiovascular disease (CVD) history.

Results:

VA and non-VA participants had similar diabetes duration (4.0 years), HbA1c (7.5%), and BMI (34 kg/m²); however, VA participants had more individuals ≥65 years (37.3% vs 19.8%,p<0.001), men (90.0% vs 55.2%,p<0.001), hypertension (75.8% vs 63.6%,p<0.001), hyperlipidemia (76.6% vs 64.6%,p<0.001), current smokers (19.0% vs 12.1%,p<0.001), nephropathy (20.4% vs 17.0%,p<0.05), albuminuria (18.4% vs 15.1%,p<0.05), and CVD (10.4% vs 5.2%,p<0.001). In those without CVD, more VA participants were treated with lipid (70.8% vs 59.5%, p<0.001) and blood pressure (74.9% vs 65.4%, p<0.001) lowering medications, and had LDL-C<70 mg/dl (32.9% vs 24.2%,p<0.05). Among those with CVD, more VA participants had BP<140/90 (80.2% vs 70.1%,p<0.05) after adjusting for demographics.

Conclusion:

GRADE participants at VA sites had more T2DM complications, greater CVD risk and were more likely to be treated with medications to reduce it, leading to more LDL-C at goal than non-VA participants, highlighting differences in diabetes populations and care.

1. Introduction

More than 30 million Americans have diabetes, of which 90% to 95% have type 2 diabetes (T2D) [1]. Type 2 diabetes is heterogeneous in terms of both pathophysiology [2, 3] and treatment [4–7]. The U.S. Department of Veterans Affairs (VA) reports that nearly one in four men and women who served in the US armed forces have diabetes [8, 9]. Roughly 18 million individuals, or 7% of the US population, were veterans of the US Armed Forces in 2018 [10]. Nine million receive care in the VA through 170 VA Medical Centers and more than 1000 outpatient clinics [11]. Compared to the general population, veterans have higher rates of obesity [12], tend to be older, have lower income, and have limited access to high-quality, healthy food. Each of these characteristics and social determinants of health are associated with greater diabetes risk and an increased frequency of diabetes-related complications. Conversely, by providing standard care to all veterans, the VA system has reduced many racial and ethnic health care disparities[13].

To standardize care, the VA system has employed a variety of quality improvement initiatives and collaboratives, as well as innovative programs. As an integrated health care system, the VA has implemented several simultaneous, national-level quality of care improvement strategies including leveraging a single electronic health record (EHR), unified nationwide guidelines, and effective performance monitoring. With respect to diabetes, the VA has specific diabetes care guidelines that are aligned with other national guidelines and has access to extensive clinical decision support tools [14] along with the implementation of the patient-aligned care team (PACT) primary care model and other initiatives such as adoption of the Choosing Wisely campaign to avoid unnecessary medical testing [15, 16].

The Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE) is a 36-clinical center national comparative effectiveness trial which was conducted to identify the most effective second glucose-lowering drug when added to metformin, for T2D over time. Ten of the 36 clinical centers were based at VA Medical Centers or had VA subsites, while the other sites represented a mix of academic, community, and health maintenance organization practices. All participants underwent identical screening processes and met identical eligibility criteria. Nevertheless, we seek to evaluate if there were substantial population differences comparing VA to non-VA sites. We also assessed differences in medical management at baseline since the GRADE cohort offered an opportunity to examine a snapshot of current diabetes care across different care settings. Specifically, this paper aims to evaluate whether GRADE participants’ baseline cardiovascular risk factors and risk factor management differed between participants enrolled at VA and non-VA clinical sites, and according to sociodemographic factors. We hypothesized that, after adjusting for baseline demographic differences, given quality improvement efforts, participants enrolled at VA sites would be more likely to be treated according to established guidelines and to meet standard diabetes care parameters than non-VA site participants.

2. Methods

2.1. Study design/setting/participants

We performed a cross-sectional comparison of baseline characteristics and rates of diabetes care target achievements in participants enrolled at VA and non-VA sites in the GRADE study, including the percentages of those with hemoglobin A1c (HbA1c)<7% (<53 mmol/mol), blood pressure (BP) <140/90 mmHg, low-density lipoprotein cholesterol (LDL-C) <100 mg/dl (<2.8 mmol/l) and <70 mg/dl (<1.8 mmol/l), and for those with CVD, treatment with statin and renin-angiotensin-aldosterone system (RAAS) BP medications, including an angiotensin converting enzyme inhibitor (ACEi) or angiotensin receptor blocker (ARB).

The GRADE study design has been previously described [17]. Briefly, participants had T2D for less than 10 years, and were diagnosed at age >30 years in non-American Indian (AI) / Alaska Native (AN) or age > 20 years for AI/AN, taking metformin monotherapy (at least 1000 mg/day), HbA1c 6.8%–8.5% (51–70 mmol/mol) at randomization, and willing to be randomly assigned to take a second glucose-lowering medication, including insulin injections, to maintain HbA1c per study protocol. Metformin was titrated to the maximally tolerated dose, up to 2000 mg daily, during run-in. The study was conducted at 36 funded clinical centers which included 9 additional sub-sites across the US with a uniform approach to management. Ten sites were based at VA Medical Centers, while the other centers represented a mix of academic, community, and health maintenance organization (HMO) practices.

2.2. Variables/Assessments/Data sources

VA sites were defined as those based at VA hospitals or clinics; 10 the 36 GRADE clinical centers were based at VA hospitals or clinics or had a subsite at a VA hospital or clinic (see list of sites and centers, Appendix). VA sites mostly enrolled veterans who received their usual care at that medical center, but some VA sites also recruited members of the local community who were not veterans and did not receive their usual medical care at the VA site. Specifically 90 non-veterans recruited from outside the VA system enrolled at VA sites (total VA site enrollment n=1216). In addition, four non-VA sites enrolled 101 veterans (of total non-VA site enrollment n=3,831), who may have received some care in the VA system. Therefore, 93% or more of participants enrolled at VA sites were veterans and less than 3% of participants enrolled at non-VA sites were veterans who may have received some or all of their care at VA hospitals or clinics (Appendix Table A.1).

Participant baseline characteristics including race and ethnicity, medical history, current medications, alcohol intake, smoking status, and educational attainment were self-reported and obtained through interviews conducted by research staff. History of hypertension, hyperlipidemia, heart attack or stroke, and retinopathy were obtained by self-report. CVD was defined as history of myocardial infarction or stroke. Nephropathy was defined as moderately or severely increased albuminuria (≥30 mg/g or >300 mg/g albumin:creatinine, respectively) or eGFR<60 ml/min/1.73m²at baseline. All physical and metabolic measurements, as well as electrocardiogram (ECG) assessment, were obtained by centrally trained, certified staff. Height, weight, and blood pressure were taken in duplicate. Height was recorded to the nearest 0.1 cm and weight to the nearest 0.1 kg. Seated blood pressure was taken after resting for five minutes and repeated after one minute; measurements were averaged. Diabetic peripheral neuropathy (DPN) was assessed annually with the modified Michigan Neuropathy Screening Instrument (MNSI) that included the 15-item interviewer-administered symptom questionnaire and a bilateral lower extremity clinical examination assessing ankle reflexes and vibration sensation at the great toes. DPN was defined based on the MNSI score ≥ 7 or an examination score ≥2.5 as previously described [18]. All laboratory tests were performed by the Central Biochemistry Laboratory (Advanced Research and Diagnostic Laboratory, Department of Laboratory Medicine and Pathology, at the University of Minnesota) using standardized laboratory procedures. HbA1c in GRADE is standardized per NGSP protocol [19]. Baseline physical assessment and laboratory values are reported; laboratory values were obtained over a period of 6 weeks at the final run-in visit or at randomization.

2.3. Statistical methods

We analyzed the baseline characteristics and rates of meeting diabetes care parameter targets by enrollment at clinical sites (VA site or non-VA site). Continuous variables were checked for symmetry and then summarized as means plus or minus standard deviations, and for the triglyceride variable, the median and interquartile range (IQR) is also provided. Categorical variables are summarized as counts and column percentages. History of CVD is a composite binary variable comprised of history of myocardial infarction (MI) by self-report (yes/no), or stroke by self-report (yes/no).

We evaluated prevalence of diabetes-related complications at baseline, including CVD (heart attack or stroke), retinopathy, neuropathy, nephropathy, and albuminuria. We evaluated whether differences in CVD risk factors remained after accounting for demographic factors (e.g. age, sex, race and ethnicity), by fitting a logistic regression model with history of CVD as the outcome variable, VA/non-VA clinical site as the main effects variable, and the demographic variables (age, sex, race, and ethnicity) as possible confounders. We then reported the p-value for the main effect variable, VA/non-VA clinical site.

We also compared management of CVD risk factors, stratified by presence of CVD, to evaluate differences between VA and non-VA clinical sites overall (using a Pearson’s chi-squared test with Yates’ continuity correction) and then adjusted for age, sex, race, and ethnicity. The adjustment was performed using an ANOVA likelihood ratio test with a binomial family adjusted for age, sex, race, and ethnicity. All analyses were conducted in R software (R Core Team, Vienna, Austria: 2020).

3. Results

Among the 5047 randomized GRADE participants, 24% (n=1,216) were enrolled at VA sites (Table 1). At baseline, participants at VA sites were more likely to be older (mean age 60.1 [SD 9.4] vs 56.2 [SD 10.0], p<0.001), and male (90.0% vs 55.2%, p<0.001) but were less likely to be Hispanic (9.1% vs 21.4%, p<0.001), with similar proportions of White (66.4% vs 65.4%) and Black/African American (21.6% vs 19.2%) participants. VA site participants, compared to non-VA participants, were more likely have a high school degree or greater (97.4% vs 91.3%, p<0.001) with 76.4% vs 70.8% with at least some college education and only 2.6% vs 8.7% without a high school degree (p<0.001 for overall education level comparison).

Table 1:

Baseline Characteristics by VA vs. Non-VA site

	VA	Non-VA	p-value
N (sites or centers)	10	16
N participants	1,216	3,831
Age at baseline visit (years)	60.1±9.4	56.2±10.0	<0.001
Age group (years)			<0.001
<45	78(6.4%)	545(14.2%)
45− <55	285(23.4%)	1,151(30.0%)
55− <65	399(32.8%)	1,378(36.0%)
65+	454(37.3%)	757(19.8%)
Sex			<0.001
Women	122(10.0%)	1,715(44.8%)
Men	1,094(90.0%)	2,116(55.2%)
Race			<0.001
American Indian/Alaska Native	3(0.2%)	134(3.5%)
Asian	63(5.2%)	119(3.1%)
Hawaiian/Pacific Islander	21(1.7%)	7(0.2%)
Black or African American	263(21.6%)	737(19.2%)
White	807(66.4%)	2,507(65.4%)
Other/multiple	53(4.4%)	266(6.9%)
Unknown/not reported	6(0.5%)	61(1.6%)
Ethnicity			<0.001
Hispanic/Latino	111(9.1%)	818(21.4%)
Not Hispanic/Latino	1,092(89.8%)	2,985(77.9%)
Unknown/not reported	13(1.1%)	28(0.7%)
Education completed			<0.001
<High school	32(2.6%)	332(8.7%)
High school graduate	255(21.0%)	784(20.5%)
Some college	439(36.1%)	1,024(26.7%)
College degree or above	490(40.3%)	1,690(44.1%)
Duration of diabetes (years)	4.0±2.7	4.0±2.8	0.802
Screening metformin dose (mg/day)	1599.7±518.2	1567.8±527.2	0.063
Baseline metformin dose (mg/day)	1933.4±227.0	1947.7±196.7	0.049
Medical history
Heart attack/stroke	127(10.4%)	201(5.2%)	<0.001
Retinopathy	20(1.6%)	29(0.8%)	0.010
Neuropathy	522(43.5%)	1,593(41.7%)	0.305
Nephropathy	248(20.4%)	650(17.0%)	0.007
Hypertension	922(75.8%)	2,438(63.6%)	<0.001
Hyperlipidemia	931(76.6%)	2,475(64.6%)	<0.001
Current medications
Blood pressure medications	935(76.9%)	2,560(66.8%)	<0.001
ACE inhibitor or angiotensin-receptor blocker	764(62.8%)	2,169(56.6%)	<0.001
Lipid-lowering medications	910(74.8%)	2,408(62.9%)	<0.001
Statin	882(72.5%)	2,328(60.8%)	<0.001
Aspirin	611(50.2%)	1,677(43.8%)	<0.001
Smoking status			<0.001
Never	469(38.6%)	2,266(59.1%)
Past	516(42.4%)	1,101(28.7%)
Current	231(19.0%)	464(12.1%)
Physical Measurements
Weight (kg)	103.9±20.9	98.7±22.6	<0.001
BMI (kg/m2)	34.2±6.2	34.3±7.0	0.476
Systolic (mmHg)	128.8±14.5	128.2±14.8	0.255
Diastolic (mmHg)	78.2±10.0	77.0±9.8	<0.001
Laboratory measurements
Cholesterol (mg/dL)	158.5±38.9	165.5±37.1	<0.001
Triglycerides (mg/dL)
Mean (SD)	159.9±108.0	148.5±97.2	0.001
Median (IQR)	132.0 [90.0, 193.0]	124.0 [86.0, 180.0]	0.004
HDL (mg/dL)	40.8±9.7	44.2±10.7	<0.001
LDL (mg/dL)	86.2±31.5	91.9±31.5	<0.001
HbA1c(%)	7.5±0.5	7.5±0.5	0.148
HbA1c (SI units)	58.2±5.2	58.4±5.3	0.148
Fasting glucose (mg/dL)	151.5±31.2	151.5±30.8	0.974
Fasting insulin (mU/L)	22.9±15.2	21.6±15.0	0.013
Fasting C-peptide (nmol/L)	1.4±0.6	1.3±0.5	<0.001
Moderately elevated albuminuria (30–299 mg/g)	199(16.4%)	517(13.5%)	0.014
Severely elevated albuminuria (≥300 mg/g)	24(2.0%)	60(1.6%)	0.401
Serum creatinine (mg/dL)	0.9±0.2	0.8±0.2	<0.001
Management/treatment goal
LDL <100 mg/dL	804(69.6%)	2,347(63.5%)	<0.001
LDL <70 mg/dL	401(34.7%)	920(24.9%)	<0.001
HbA1c <7% (<53 mmol/mol)	187(15.4%)	538(14.0%)	0.267
Urinary albumin to creatinine ratio <30 mg/g	991(81.6%)	3,249(84.9%)	0.007
At BP goal (<140/90 mmHg) or on ACEi/ARB	1,123(92.4%)	3,488(91.0%)	0.176
At target UACR (<30 mg/g creatinine) or on ACEi/ARB	1,153(94.8%)	3,627(94.7%)	0.983

Continuous variables are summarized as mean +/− standard deviation. For the triglyceride variable, the median and interquartile range (IQR) is also provided. Categorical variables are summarized as counts and column percentages.

Although diabetes duration, HbA1c, screening metformin dose, BMI, and systolic blood pressure were similar at baseline, VA site participants had slightly higher levels of fasting insulin, C-peptide, and triglycerides compared to non-VA site participants. In addition, individuals enrolled at VA sites were more likely to report a history of hypertension (75.8% vs 63.6%, p<0.001), hyperlipidemia (76.6% vs 64.6%, p<0.001), heart attack or stroke (10.4% vs 5.2%, p<0.001), retinopathy (1.6% vs 0.8%, p<0.05), and nephropathy (20.4% vs 17.0%, p<0.05), to have urinary albumin-to-creatinine ratio (UACR) ≥30 mg/g (18.4% vs 15.1%, p<0.05), and to be current smokers (19.0% vs 12.1%) or former smokers (42.4 vs. 28.7%. p<0.001), compared to participants at non-VA sites.

In reference to cardiovascular risk factor management, a greater proportion of VA participants reported taking blood pressure (76.9% vs 66.8%, p<0.001) and lipid-lowering (74.8% vs 62.9%, p<0.001) medications, and aspirin (50.2% vs 43.8%, p<0.001).

With regard to achievement of treatment goals, VA participants were more likely to have LDL-C<100 mg/dL (<2.6 mmol/l) compared to non-VA participants (69.6% vs 63.5%, p<0.001); VA participants were also more likely to have LDL-C<70 mg/dL (<1.8 mmol/l). Participants enrolled at VA sites were similar to those at non-VA sites with respect to having HbA1c <7% (<53 mmol/mol) (15.4% vs 14.0%), having either blood pressure <140/90 mmHg or treatment with RAAS inhibitors (92.4% vs 91.0%), and either having absence of albuminuria or receiving treatment with RAAS inhibitors (94.8% vs 94.7%). Of note, greater than 90% of GRADE participants met their blood pressure and UACR goals or reported appropriate medications for these conditions (Table 1).

Differences in the prevalence of cardiovascular disease between participants enrolled at VA and non-VA sites remained statistically significant (p=0.034) after adjusting for differences in baseline demographics (age, sex, race, and ethnicity [Hispanic/non-Hispanic]).

Among those with history of CVD and after adjusting for demographic factors (Table 2a), VA site participants had a significantly higher proportion of individuals within target BP (<140/90 mmHg) compared to non-VA participants (80.2% vs 70.1%, p<0.05). However, there were similar proportions of participants with HbA1c<7% (<53 mmol/mol) and LDL-C<100 mg/dl. No differences were observed between VA and non-VA participants with history of CVD for treatment with blood pressure lowering medications, statins, or aspirin. With regard to a more aggressive LDL-C target for individuals with history of CVD, a higher proportion of VA site participants had LDL-C<70 mg/dL (<1.8 mmol/l) compared to non-VA participants (50.0% vs 36.9%, p<0.05) overall, but this difference was no longer significant after adjusting for age, sex, race, and ethnicity.

Table 2a:

Diabetes care targets, including cardiovascular risk factor management in participants with prior history of cardiovascular disease at VA sites vs. non-VA sites

	VA	Non VA	p-value	Adj. p-value*
N	127	201
Among participants With CVD history
HbA1c <7% (<53 mmol/mol)	23(18.1%)	22(10.9%)	0.094	0.093
BP <140/90 mmHg	101(80.2%)	141(70.1%)	0.060	0.035
Treated for HTN	119(93.7%)	187(93.0%)	0.993	0.721
LDLc <70 mg/dL (1.8 mmol/L)	61(50.0%)	72(36.9%)	0.029	0.340
LDLc <100 mg/dL (2.6 mmol/L)	99(81.1%)	145(74.4%)	0.208	0.700
On statin	111(87.4%)	169(84.1%)	0.504	0.872
Aspirin	104(81.9%)	154(76.6%)	0.319	0.867

Counts and column percentages of participants with CVD history with key diabetes management treatment goals at VA and non-VA clinical sites.

^*Adjusted for age, race, ethnicity, and sex

For individuals without a history of CVD and after adjusting for demographic factors (Table 2b), more VA clinical site participants had LDLC-<70 mg/dl (<1.8 mmol/l) (32.9% vs 24.2%, p<0.05), and reported taking blood-pressure lowering medications (74.9% vs 65.4%, p<0.05), statins (70.8% vs 59.5%, p<0.05), and aspirin (46.6% vs 42.0%, p<0.05), when compared to non-VA participants. In contrast, similar proportions of VA participants and non-VA participants had HbA1c<7% (<53 mmol/mol) and BP<140/90 mmHg. Sensitivity analyses restricted to participants aged 40 years and older and 50 and older without CVD (Appendix Tables A.2a and A2.b) showed similar results.

Table 2b:

Diabetes care targets, including cardiovascular risk factor management in participants without prior history of cardiovascular disease at VA sites vs. non-VA sites

	VA	Non VA	p-value	Adj. p-value*
N	1,089	3,630
Among participants Without CVD history
HbA1c <7% (<53 mmol/mol)	164(15.1%)	516(14.2%)	0.518	0.935
BP <140/90 mmHg	801(73.6%)	2,757(76.0%)	0.113	0.467
Treated for HTN	816(74.9%)	2,373(65.4%)	<0.001	0.006
LDLc <70 mg/dL (1.8 mmol/L)	340(32.9%)	848(24.2%)	<0.001	0.045
LDLc <100 mg/dL (2.6 mmol/L)	705(68.2%)	2,203(62.9%)	0.002	0.779
On statin	771(70.8%)	2,159(59.5%)	<0.001	0.028
Aspirin	507(46.6%)	1,523(42.0%)	0.008	0.003

Counts and column percentages of participants without CVD history with key diabetes management treatment goals at VA and non-VA clinical sites.

^*Adjusted for age, race, ethnicity, and sex

4. Discussion

We found substantial differences in baseline characteristics of GRADE participants with type 2 diabetes of less than 10 years duration, treated with metformin alone, and stratified by enrollment at VA and non-VA clinical sites. These reflect differences in the background population and management protocols used at these two categories of clinical centers. Participants enrolled at VA sites were older and more likely to be male, and less likely to be Hispanic or to have less than a high school education compared to non-VA participants. While participants at VA sites had similar glycemic and weight parameters to those in non-VA sites, VA participants had higher rates of hypertension, hyperlipidemia, and cardiovascular disease, suggestive of a more insulin resistant phenotype. They had higher levels of insulin and C-peptide for similar baseline glucose levels. They were also more likely to have smoked in the past or currently, which increases their CVD risk.

These findings have implications for clinical care and clinical research. First, with early adoption of computerized electronic health records, many reports of quality initiatives have emerged from the VA System, with prior reports suggesting a higher quality of diabetes care in the VA compared to commercial managed care [20]. It is thus important to understand the ways in which the VA diabetes population may differ from the non-VA diabetes population in large-scale studies such as GRADE. Second, these findings are important for interpretation of clinical research. GRADE, like many other trials, recruited a substantial proportion of its cohort from the VA. Yet, VA participants systematically differ from non-VA participants, again, with implications for the interpretation of the overall generalizability of trial results.

In terms of clinical care, participants at VA sites had more indications for CV risk factor management, and those with CVD were more likely to have blood pressure at goal compared to those enrolled at non-VA sites. VA site participants without CVD were more likely to be treated with blood pressure medication and statins and more likely to have LDL-C<70 mg/dl than those at non-VA sites, although blood pressure control and HbA1c<7% (<53 mmol/mol) were similar across sites. A greater proportion of veterans without CVD had LDL-C<70 mg/dl, even after adjusting for demographic factors and performing age-restricted sensitivity analyses. These analyses suggest that VA participants were more likely to have guideline-concordant CVD risk factor management, measured by both process (ACEi/ARB, statin use) and outcome (blood pressure, LDL-C) measures.

There are a few studies that have compared diabetes care delivered at the VA to that delivered in other health care organizations. An analysis using the 2000 BRFSS data showed that male VA patients with diabetes received preventive care services at equivalent or higher levels than their counterparts receiving care outside the VA and non-veterans [21]. An analysis of respondents to the 2003 BRFSS showed that veterans were more likely to have foot and dilated eye exams, have aspirin treatment, and receive influenza and pneumococcal vaccines [22]. In a cross-sectional patient survey and retrospective review of medical records, VA Medical Centers were compared to managed care organizations in matched geographic regions [20]. Patients in the VA system had better scores on process measures (annual HbA1c and eye exams), better LDL-C (52% vs. 36% with LDLc<100 mg/dl) (<2.8 mmol/l), and similar poor blood pressure control (only 52% to 53% with BP<140/90 mmHg). The satisfaction scores were similar, although VA patients were slightly more satisfied with overall quality of diabetes care. A more recent comparison of quality of care conducted by the RAND corporation using HEDIS measures for outpatient effectiveness showed better diabetes care with HbA1c<9% (<53 mmol/mol), BP<140/90, and LDL-C<100 mmHg) (<2.8 mmol/l) in VA facilities than in commercial, Medicare, and Medicaid HMOs [23]. Similarly, VA patients were more likely than commercially insured patients to meet recommended care targets [20]. The VA’s long history of EHR innovation in population management may be one factor contributing to these outcomes [24, 25].

Phenotypic heterogeneity within type 2 diabetes according to insulin sensitivity and insulin resistance is increasingly recognized, including within the GRADE study [2, 3, 26]. It is notable that despite equivalent baseline fasting glucose, HbA1c, and BMI, the older, male participants at VA sites were more likely to have slightly but statistically significantly higher insulin and C-peptide levels as well as higher rates of nephropathy and elevated albuminuria, on average, compared to the younger, more gender-balanced population enrolled at non-VA sites. This physiologic signal, while relatively weak, is consistent with the higher prevalence of CVD and CVD risk factors.

This sample has several unique characteristics that affect interpretation of the results. It was restricted to individuals with type 2 diabetes who met inclusion criteria for the GRADE trial, limiting variability across glycemia and glycemic management. Moreover, this is not a population-based study, since all participants elected to enroll in a clinical trial. However, in contrast to previous studies of veterans and non-veterans, there were no differences in the methodology used to identify the comparison groups, which may have introduced selection biases. Most VA and non-VA sites were in large cities. VA sites were high-complexity sites offering specialty care, and many of the non-VA sites were academic medical centers. It is known that veterans at rural centers have slightly but statistically significantly worse glycemic management than those at urban centers [27]. A recent report from a national cohort of veterans with diabetes showed geographic and racial/ethnic differences in management, which were not explained by differences in comorbidity burden, medication adherence, access, and health care utilization[27]. There is some misclassification of veterans and those exposed to VA protocols by VA or non-VA site, since a small number of non-veterans were enrolled at VA sites, and 3% or less of participants at non-VA sites were veterans who may have been exposed to VA management protocols. This would tend to minimize differences between groups. Yet the vast majority (>90%) of participants at VA sites were veterans who received at least part of their care in the VA system. These findings are consistent with the hypothesis that quality of care improvement efforts and equitable access to affordable medication across the VA system increase the likelihood that CVD and related risk factors will be intensively treated in the appropriate populations.

In conclusion, among GRADE participants with type 2 diabetes with HbA1c 6.8–8.5% (50.8–69.4 mmol/mol) treated with metformin, those enrolled at VA sites were older, more likely to be male, had higher rates of cardiovascular risk factors and complications (cardiovascular disease, nephropathy) at baseline. Those enrolled at VA sites were more likely to report taking medications to reduce their CVD risk compared to those enrolled at non-VA sites. VA site participants were more likely to meet LDL cholesterol targets and other management goals. These results highlight differences in population characteristics and care patterns seen between patients receiving care in VA and non-VA health care organizations, suggesting that comprehensive medical coverage, uniform care protocols and quality initiatives such as those available at the VA may move the needle on diabetes care parameters, even in a higher risk population. In addition, understanding the difference in VA and non-VA clinical diabetes populations may aid in interpretation of the generalizability of results of clinical research interventions. Future efforts will examine whether these baseline differences in clinical characteristics and care will impact outcomes and response to diabetes therapy.

Highlights.

• GRADE enrolled participants with type 2 diabetes treated with metformin enrolled at VA and non-VA clinical centers.

• Participants enrolled at VA and non-VA centers had different baseline characteristics and risk factor management.

• These differences reflected the background population and management protocols prevalent at these two categories of clinical centers.

• VA participants had similar HbA1c and weight as non-VA participants.

• VA participants had higher rates of hypertension, hyperlipidemia, and cardiovascular disease compared to non-VA participants.

• VA site participants were generally more likely than non-VA participants to have guideline-concordant CVD risk factor management.