Demographic Predictors of Elevated LDL Cholesterol in Type 1 Diabetes: A Cross-Sectional Study Free

We conducted a cross-sectional analysis of individuals with type 1 diabetes using the Premier Healthcare Database (PHD). PHD is a large, U.S. hospital-system–based, all-payer electronic database containing information originating in 2000 from >315 million patients in >1,322 hospitals and health systems across the United States. The study was granted University of Virginia Human Subjects Research Institutional Review Board exemption (submission #301153).

Individuals with type 1 diabetes, defined based on International Classification of Diseases, 10th revision (ICD-10), diagnosis codes, who were 55–75 years of age and had outpatient encounters between 1 January 2020 and 31 December 2022 were included. We selected the age range of 55–75 years to examine only patients who met criteria for cholesterol-lowering treatment and to avoid the potential influence of pregnancy or breastfeeding on cholesterol-lowering practices (20).

We excluded individuals with inpatient or emergency department encounters to avoid acute CVD, in which cases more aggressive lipid targets would not yet be implemented. Individuals with missing LDL cholesterol data or incomplete demographic data were excluded (Supplementary Figure S1).

The outcome variable was presence of elevated LDL cholesterol (≥100 mg/dL). LDL cholesterol values recorded in SI units (i.e., mmol/L) or accompanied by a reference range incompatible with LDL cholesterol values (e.g., reference range <40 mg/dL, suggesting mislabeling of LDL cholesterol data) were excluded. If duplicate LDL cholesterol values were present for a single individual, the lowest value was accepted. This process changed the LDL cholesterol target classification of 16 individuals (0.3% of the study population). Predictor variables examined included age, sex, race, ethnicity, CVD diagnosis (based on the ICD-10 codes listed in Supplementary Table S1), and insurance payer (Medicare/other non-Medicaid government payer, Medicaid, commercial, managed care, self-pay, or other).

A multiple logistic regression model was used to determine which of the aforementioned predictor variables were uniquely associated with LDL cholesterol ≥100 mg/dL after considering the LDL classification information explained by the other predictor variables.

We performed a secondary analysis restricted to patients with a CVD diagnosis because of the lower recommended LDL cholesterol target (<70 mg/dL) in this population during the observation period (9–12). The outcome variable was presence of LDL cholesterol over target (≥70 mg/dL). Predictor variables examined included age, sex, race, ethnicity, and insurance payer. A multiple logistic regression model was again used to determine which of the aforementioned predictor variables were uniquely associated with LDL cholesterol ≥70 mg/dL after considering the LDL cholesterol classification information explained by the other predictor variables.

The data that support the findings of this study are available from Premier Healthcare Solutions, Inc., but restrictions apply to the availability of these data, which were used under license for the current study and therefore are not publicly available. The analytic code used in the current study is available from the corresponding author upon reasonable request. Data are nontransferable in accordance with a data use agreement with Premier Healthcare Solutions, Inc.

Of the 173,178 outpatients in the PHD with type 1 diabetes, 5,344 had LDL cholesterol data. (Descriptive characteristics of full outpatient cohort are available in Supplementary Table S2.) Of these, 4,987 patients also had complete demographics data, including known race and ethnicity, and these were included in the final analysis. Cohort descriptive statistics are provided in Table 1. The cohort was 52% female, predominately White (89.8%), and of non-Hispanic ethnicity (96.7%), with a mean age of 65.4 ± 5.9 years. The most common insurance payer was Medicare or other non-Medicaid government payer (62.8%). LDL cholesterol was ≥100 mg/dL in 1,742 (35%) of these individuals. The LDL cholesterol frequency distribution for the cohort is shown in Figure 1A.

In the multiple logistic regression model examining factors associated with LDL cholesterol ≥100 mg/dL, several predictor variables were independently associated with elevated LDL cholesterol (Table 2 and Table 3). Female sex (adjusted odds ratio [aOR] 1.61, 95% CI 1.43–1.81, P <0.001) and Medicaid insurance payer were each independently associated with LDL cholesterol above target. Medicaid payer status increased the odds of having LDL cholesterol above goal compared with commercial insurance (aOR 1.92, 95% CI 1.30–2.84, P = 0.001), Medicare or other non-Medicaid governmental insurance (aOR 1.37, 95% CI 1.11–1.55, P = 0.008), or managed care (aOR 1.56, 95% CI 1.12–2.16, P = 0.008), but not self-pay. Conversely, CVD diagnosis and older age were associated with reduced odds of having LDL cholesterol ≥100 mg/dL. The odds of LDL cholesterol reaching the target increased by 8% per 5 years of age in this cohort between the ages of 55 and 75 years. The model as a whole provided significant information about above-target LDL cholesterol (P <0.001). There were no significant interactions between sex and any other predictor variables, and the Hosmer-Lemeshow goodness-of-fit test indicated no evidence of lack of fit (P = 0.9), indicating satisfactory model fit.

In the subgroup of individuals with diagnosed CVD (n = 321), 149 (46%) were female, 20 (6.2%) were Black, 20 (6.2%) were of other race, and 17 (5.3%) identified as Hispanic ethnicity. A majority of individuals with CVD (69.5%) were insured by Medicare, and 13 (4%) were insured by Medicaid. In these patients with established CVD, 174 (54%) had an LDL cholesterol ≥70 mg/dL (Figure 1B). Thus, more than half of these individuals at very high CVD risk had LDL cholesterol levels above target. Using current guideline recommendations that individuals with CVD should have an LDL cholesterol target of <55 mg/dL, only 79 patients (24.6%) attained this goal. A frequency table by sex and LDL cholesterol ≥70 mg/dL (Supplementary Table S3) indicates that 60.4% (90 of 149) of females compared with 48% (84 of 172) of males did not meet LDL cholesterol recommendations, which was statistically significant using a Fisher exact test (P = 0.043).

After adjusting for other covariates in the multiple logistic regression model in the CVD subgroup analysis, female sex was not significantly associated with LDL cholesterol above target (aOR 1.47, 95% CI 0.92–2.36, P = 0.110), nor were the other covariates (Supplementary Table S4 and Supplementary Table S5). This study was likely underpowered to find statistical significance due to sex in the subgroup with CVD. If designing a new independent study, we would predict that 950 subjects (509 females and 441 males) with CVD would be required to have 0.80 statistical power to detect the sex-associated aOR of 1.47 observed in the current subgroup of patients with a type I error rate of 0.05. Details of the sample size calculation are provided in the Supplementary Material.

This cross-sectional, real-world study has several important findings. First, many (35%) of the individuals with type 1 diabetes in this cohort had elevated LDL cholesterol. Second, a preponderance of individuals with established coronary atherosclerosis, cerebral infarction, or peripheral artery disease did not reach the LDL cholesterol target of <70 mg/dL, which is recommended for high-risk individuals, during the study observation period. These findings raise alarm that preventable cardiovascular risk factors are not being adequately mitigated in people with type 1 diabetes.

After adjusting for CVD and other predictor variables such as age and insurance payer, female sex was associated with 61% higher odds of not attaining an LDL cholesterol level <100 mg/dL. This finding was in line with our hypothesis that females would be less likely to reach treatment goals. Medicaid insurance coverage was associated with 32–97% higher odds of having an LDL cholesterol level above the treatment target compared with other insurance payers. Conversely, prior CVD diagnosis was associated with a greater odds of having an LDL cholesterol level <100 mg/dL, and older age increased the odds of LDL cholesterol reaching the target despite the recommendation that all patients ≥40 years of age with type 1 diabetes receive aggressive LDL cholesterol–lowering treatment. Contrary to our hypothesis, we did not find that individuals of Black race or Hispanic ethnicity were more likely to have above-target LDL cholesterol levels after adjusting for other demographic factors. Failure to detect significance could have been related to the poor diversity in this cohort, which was predominately White and non-Hispanic.

The finding that women in the type 1 diabetes population were less likely to reach LDL cholesterol goals compared with their male counterparts is novel. We predicted this outcome based on previously elucidated sex-based inequities in terms of CVD risk assessment and statin prescribing practices in other populations. In a primary care setting, women are less likely than men to have their CVD risk assessed (21). For women with or at high risk for CVD, prescription of preventive medications such as statins occurs less frequent than for men (22–24). In the DCCT intervention and subsequent EDIC observational study, women with type 1 diabetes were less likely than their male counterparts to report statin use, although LDL cholesterol levels were similar between sexes (6).

This study was not designed to determine causality for the higher LDL cholesterol levels. Several provider, patient, and socioeconomic factors could explain the failure to reach target LDL cholesterol levels. Potential provider factors include a knowledge gap regarding patients with type 1 diabetes being at high CVD risk or implicit bias regarding age, sex, or socioeconomic status. For example, relatively young female patients may not phenotypically appear to be at high CVD risk. Patient factors could include statin intolerance, perceived intolerance, or nonacceptance of statins. Women are more likely than men to report statin-related side effects and to stop or switch their statin because of side effects and are less likely to believe that statins are safe and effective (21,25). However, statins are clearly safe in women (26). Women are also less likely to feel educated by their provider on their risk of CVD compared with men, which could contribute to discontinuing statin treatment (21). Medicaid insurance as a risk factor for higher LDL cholesterol is likely a surrogate for low income status. Socioeconomic factors may be a barrier to optimal treatment because of prescription costs and limited access to care for routine monitoring and to obtain medication prescriptions. Further study is needed to define causes of sex-, age-, and insurance/ socioeconomic–based inequities in elevated LDL cholesterol in people with type 1 diabetes.

To combat clinical inertia and nonadherence, clinicians can take actions such as setting electronic medical record reminders for annual cholesterol screenings to assess treatment responses and adherence. Uncovering elevated LDL cholesterol levels could facilitate a discussion of treatment barriers such as intolerance or affordability issues. Providers should consider alternative or additional LDL cholesterol–lowering agents such as proprotein convertase subtilisin/kexin type 9 inhibitors for individuals who are statin intolerant or not meeting LDL cholesterol targets. Additionally, the low proportion of individuals reaching LDL cholesterol targets in this study underscores the importance of discussing cardiovascular risk with patients who have type 1 diabetes to promote acceptance of cardiovascular risk reduction therapy. This effort may be particularly important in certain groups at higher odds of having elevated LDL cholesterol, such as women.

This study has several limitations. First, reliable prescription fill data for statins or other lipid-lowering medications were not available for this study, but this information would further clarify cardiovascular risk reduction practices. In a study of the PINNACLE outpatient registry cohort, 53% of patients with coronary artery disease who were not treated with a statin had an LDL cholesterol level <100 mg/dL (27). Nevertheless, statin therapy lowers cardiovascular risk even in individuals with LDL cholesterol levels <70 mg/dL (17).

Second, this study was cross-sectional, but longitudinal LDL cholesterol data would better define whether patients not reaching the LDL cholesterol target of <100 mg/dL experienced a 30–49% LDL cholesterol reduction, considered an acceptable response to moderate-intensity statin treatment based on major guideline recommendations (9–12). Still, there is a clear linear relationship between LDL cholesterol and CVD events, with lowest LDL cholesterol levels providing optimal benefit. In a meta-analysis of 21 statin trials, each 39 mg/dL decrease in LDL cholesterol corresponded with a 21% reduction in major vascular events (17).

Third, most of the patients with type 1 diabetes in the outpatient PHD cohort did not have cholesterol data. This lack of data could have introduced selection bias. For example, this study would fail to detect patients not meeting cholesterol targets because of a lack of access to laboratory testing or outpatient visits. In particular, we observed that racial and ethnic minorities, as well as individuals with Medicaid and commercial insurance, were modestly underrepresented in the LDL cholesterol study population compared with those with any outpatient data. This underrepresentation could have attenuated the effect size in these underrepresented groups. To validate and expand on this study’s findings, an ideal future study would include prescriptions written and prescription fill data; accurate diagnostic codes to allow more comprehensive comorbidity covariate inclusion; data on duration of diabetes; longitudinal encounter data, including frequency of ambulatory and subspeciality visits; a larger sample size of people with a CVD diagnosis; and a more racially and ethnically diverse population. Qualitative and quantitative survey research involving providers and patients is also a potential next step to identify perceptions and inform causality.

A large portion of individuals with type 1 diabetes who are ≥55 years of age do not meet LDL cholesterol treatment targets for primary or secondary cardiovascular prevention. Demographic factors independently associated with elevated LDL cholesterol include female sex, Medicaid insurance, and younger age. Further investigation to uncover the etiology of these disparities is needed.

The authors thank Dr. William B. Horton, of the University of Virginia, in Charlottesville, VA, for his guidance on data acquisition.

This work was supported by National Institutes of Health (NIH) grant K23DK131327-01A1 (to K.M.L.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

No potential conflicts of interest relevant to this article were reported.

M.E.D. and K.M.L. devised the hypothesis and wrote the first draft of the manuscript. J.T.P. performed statistical analysis, contributed to statistical methods and results, and reviewed and edited the manuscript. K.M.L. cleaned the data. K.M.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Results from this study were presented in abstract form at the American Diabetes Association’s 84th Scientific Sessions in Orlando, FL, on 22 June 2024.