Abstract
Objective To estimate the prevalence of dyslipidemia and to describe its management in Canadian primary care.
Design Retrospective cohort study using primary care electronic medical record data.
Setting Canada.
Participants Adults aged 40 years or older who saw a Canadian Primary Care Sentinel Surveillance Network contributor between January 1, 2018, and December 31, 2019.
Main outcome measures Presence or absence of dyslipidemia as identified by a validated case definition and the treatment status of patients identified as having dyslipidemia based on having been prescribed a lipid-lowering agent (LLA).
Results In total, 50.0% of the 773,081 patients 40 years of age or older who had had a primary care visit in 2018 or 2019 were identified as having dyslipidemia. Dyslipidemia was more prevalent in patients 65 or older (61.5%), in males (56.7%) versus females (44.7%), and in those living in urban areas (50.0%) versus rural areas (45.2%). In patients with documented dyslipidemia, 42.8% had evidence of treatment with an LLA. Stratifying patients by Framingham risk score revealed that those in the high-risk category were more likely to have been prescribed an LLA (65.0%) compared with those in the intermediate-risk group (48.7%) or the low-risk group (22.8%). The strongest determinants of receiving LLA treatment for dyslipidemia include sex, with males being 1.95 times more likely to have been treated compared with females (95% CI 1.91 to 1.98; P<.0001); and body mass index, with those with obesity having a significantly increased likelihood of being treated with an LLA (adjusted odds ratio of 1.36, 95% CI 1.32 to 1.41; P<.0001).
Conclusion This study provides an updated look at the prevalence and treatment of dyslipidemia among Canadians. Half of patients aged 40 years or older have dyslipidemia, with an even higher prevalence observed among adults aged 65 years or older, males, and those with obesity or other chronic conditions. There are still gaps in treatment among those with documented dyslipidemia, principally among those calculated to have high or intermediate Framingham risk scores. Particular attention should also be paid to those at higher risk for not receiving treatment, including female patients and those within normal body mass index ranges.
Cardiovascular disease (CVD) is the second leading cause of death in Canada,1 meaning there is ongoing need for effective primary and secondary prevention within primary care to improve CVD outcomes.2-4 Dyslipidemia is an important modifiable risk factor and refers to various lipid profile abnormalities, including elevated levels of low-density lipoprotein cholesterol (LDL-C) in the bloodstream.2,4-8 It is believed that high levels of LDL-C cause damage to endothelial cells, which are responsible for separating the exceedingly thrombogenic internal elastic lamina and basement membrane from various circulating components of the bloodstream.9,10 This damage ultimately leads to an inflammatory response and triggers the development of atherosclerotic plaques within blood vessels.5,7,11
The 2021 Canadian Cardiovascular Society guidelines recommend screening all individuals age 40 or older for dyslipidemia, as well as patients of any age with clinical conditions that increase their CVD risk profile.4 Although targets for lipid-lowering therapy vary worldwide, studies have shown that for every 1-mmol/L decrease in LDL-C level there is an associated 23% relative reduction in risk of major vascular events.2-4,7,11 Given the relative nature of the risk reduction, treatment guidelines follow a tiered approach, separating individuals into high-, intermediate-, and low-risk status.3,4,7 In Canada, the Framingham risk score (FRS) is commonly used for stratification and CVD risk estimation purposes. Based on FRS calculations, statin therapy should be started for individuals at high risk (≥20% 10-year risk) and for those at intermediate risk (10%-19.9%) of CVD.4,12 Additionally, there are some conditions where statin therapy is indicated independent of cholesterol levels, such as with clinical atherosclerosis (ie, myocardial infarction, acute coronary syndrome, angina, coronary artery disease by angiography, stroke, transient ischemic attack, documented carotid disease, peripheral arterial disease, claudication, and ankle brachial index value <0.9), abdominal aortic aneurysm (AAA), most cases of diabetes (age ≥40, age ≥30, and ≥15-year duration of microvascular disease), chronic kidney disease (CKD) if older than 50, and genetic dyslipidemia.4 Individuals at low risk of CVD typically do not require pharmacotherapy.4
The primary pharmacologic approach for treating dyslipidemia is a lipid-lowering agent (LLA), which primary care providers commonly prescribe. Several studies have shown that dyslipidemia is undertreated both worldwide and in Canada.13-18 There is also emerging evidence of inequitable pharmacologic management of dyslipidemia in populations including women, those with lower socioeconomic status, and Indigenous people.17,9-21 Understanding which individuals are at higher risk for undertreatment of dyslipidemia is essential to support quality improvement and to improve primary and secondary prevention of CVD outcomes.
This study sought to describe a cohort of patients with dyslipidemia and to evaluate the demographic and clinical predictors of prevention of CVD with LLA use. The main objective was to provide primary care providers and policy-makers insights into areas for improvement.
METHODS
Study design and data source
We conducted a retrospective cohort study of adults diagnosed with dyslipidemia using electronic medical record (EMR) data from the Canadian Primary Care Sentinel Surveillance Network (CPCSSN), which is a national network of practice-based research and learning networks that collects and aggregates EMR data into a large repository that can be used for practice feedback, quality improvement, surveillance, and research. Over the past decade, CPCSSN has grown to include more than 1500 primary care providers across Canada who contribute clinical data on more than 1.5 million patients. It contains a broad scope of clinical data, including diagnoses, medications, laboratory results, risk factors, and procedures. The Canadian Primary Care Sentinel Surveillance Network uses validated case definitions to identify patients with common chronic conditions, including diabetes and hypertension.21 The amount of historical clinical data on each patient is highly variable. This study included clinical data from patients in British Columbia (n=41,706), Alberta (n=176,213), Manitoba (n=100,241), Ontario (n=423,515), Quebec (n=23,163), Nova Scotia (n=5323), and Newfoundland and Labrador (n=2922).
Dyslipidemia case definition
A validated case definition was developed to identify patients with dyslipidemia. Patients met the case definition (described in supplementary material, available from CFPlus*) if they had been flagged by any or all of the following components: dyslipidemia ICD-9 code 272, lipid-lowering medication use, or abnormal lipid levels in laboratory tests (LDL-C ≥3.5 mmol/L or a total cholesterol to high-density lipoprotein cholesterol [HDL-C] ratio ≥4.3). This case-finding algorithm was developed by a review of previously published literature for dyslipidemia case definitions, an evaluation of the current Canadian guidelines for dyslipidemia screening, and an appraisal by an expert panel consisting of a family doctor (D.B.) and an endocrinologist. In contrast with previous algorithms,22-24 we did not classify patients with diabetes as having dyslipidemia if the only evidence of a lipid disorder was the prescription of an LLA. This was due to recent guidelines that recommend prescribing an LLA to all patients with diabetes, regardless of lipid levels.4 Consequently, in patients with diabetes, prescription of an LLA alone is insufficient evidence to classify a patient as having dyslipidemia.
To validate the developed case definition a manual chart abstraction was conducted that verified the dyslipidemia status of a sample of patients drawn from 1 interprofessional primary care clinic in Ontario consisting of 25 primary care providers. A minimum sample size was determined by setting the expected prevalence of dyslipidemia at 28%, a conservative estimate based on previous studies that had found prevalence ranged from 12.1% to 44.6% in adults.6,22-24 Using this prevalence, along with a 95% confidence interval and aiming to achieve a margin of error of less than 10%, 222 charts from a primary care clinic were randomly selected for review. With the algorithmic diagnosis of cases masked, a nurse reviewed the entire chart including visit notes, diagnoses, laboratory results, and medications for evidence of the presence or absence of dyslipidemia. When the nurse was uncertain about the diagnosis, an epidemiologist (R.M.) and a physician (D.B.) from the study team reviewed the chart.
The validation results were summarized using a 2 × 2 table comparing the algorithm (either a case or a noncase) with the chart review diagnosis (either a case or a noncase). Sensitivity, specificity, and positive and negative predictive values were calculated with 95% confidence intervals.
Prevalence
The algorithm was applied to the CPCSSN database (national repository) to evaluate the prevalence of dyslipidemia in adults aged 40 years or older. This age group was chosen to reduce ascertainment bias as cholesterol testing in those younger than 40 is uncommon. The population at risk was defined as the practice population in 2019, estimated using a 2-year contact group (any individual with a primary care visit in 2018 or 2019). Using a 2-year contact group has been established as the most effective methodology to capture active patients who receive most of their care from the same primary care clinic.25 Defining a practice population based on all patients with at least 1 visit within a 2- or 3-year period has been shown, through internal evaluations undertaken by CPCSSN as well as reports from other primary care research centres, to be a suitable choice for chronic disease epidemiology in Canada.25
The prevalence of dyslipidemia was evaluated in various subgroups, including by age group, sex, practice location (urban or rural), body mass index (BMI), smoking status, number of comorbid conditions, and the size of the primary care practice where a patient receives care.
A sensitivity analysis was conducted to evaluate the extent of ascertainment bias and to provide confidence in the robustness of the observed prevalence estimates. This sensitivity analysis consisted of measuring the frequency of lipid testing (count of patients who have ever had lipid testing) by age group, sex, and location.
Treatment status
A patient was considered treated if they had had at least 1 prescription for an LLA medication at any time prior to the end of the study period (January 1, 2020), which was identified using Anatomical Therapeutic Chemical (ATC) code C10, excluding bile acid sequestrants (ATC code C10AC). To avoid confounding by indication the treatment status was determined only for patients who met criterion A or B of the dyslipidemia case definition (Figure 1). Patients who met criterion A or B are referred to as patients with documented dyslipidemia.
Flow chart for study population determination
Treatment statuses of patients with documented dyslipidemia were compared with those of patients with other LLA-indicated conditions, specifically CKD, CVD, AAA, and genetic dyslipidemia. For this comparison any patient having an LLA-indicated condition was removed from the dyslipidemia group. The validated case definitions and algorithms used to classify a patient as having one of the LLA-indicated conditions can be found in supplementary material, available from CFPlus.*21
Framingham risk score
An FRS was calculated for each patient with documented dyslipidemia. The risk calculator, as provided by the Canadian Cardiovascular Society, evaluates the following risk factors: age, sex, HDL-C levels, total cholesterol levels, systolic blood pressure, smoking status, and diabetes status.26-28 Scores could be calculated only for patients with valid records for HDL-C, total cholesterol, and systolic blood pressure. Laboratory data were included in the calculation of the FRS only if the tests had been conducted prior to any LLA use (first date of any prescription for an LLA as identified using ATC code C10). Furthermore, a patient was considered a non-smoker if no smoking status had been recorded or if the patient’s status was described as having quit or having never smoked. Previous research has shown that 64.4% of patients have a smoking status recorded in the CPCSSN data repository.29 A sensitivity analysis was also undertaken where patients with no smoking data were excluded to evaluate how the proportion in each FRS category was affected by the assumption that no smoking documentation was equivalent to being a non-smoker.
Statistical analysis
Characteristics of the study cohort and subgroups were expressed as percentages and numeric values. To assess differences in the frequency of dyslipidemia across demographic and clinical factors, 2 analysis and the Fisher exact test were used.
Multivariable logistic regression was used to explore predictors of treatment with an LLA among individuals with documented dyslipidemia. Unadjusted and adjusted estimates are reported. As this was an exploratory analysis and there were no a priori hypotheses, all covariates were kept in the fully adjusted model. All data were analyzed using SAS version 9.4. A P value less than .05 was considered statistically significant.
This study was approved by the Queen’s University Health Sciences and Affiliated Teaching Hospitals Research Ethics Board (#6032985).
RESULTS
In the CPCSSN database there were 773,081 patients 40 years or older with at least 1 clinical encounter between January 1, 2018, and December 31, 2019. Of these patients, 50.0% met the case definition for dyslipidemia. Table 1 describes the proportion of patients who met any condition of the dyslipidemia case definition and summarizes the proportions of patients who met each case definition criterion (A, B, or C). The manual chart review of 222 records (Figure 2) indicated that this case definition performs well in terms of identifying patients with and without dyslipidemia, with a sensitivity of 95.4% (95% CI 87.2% to 98.6%), a specificity of 100% (95% CI 97.5% to 100%), a negative predictive value of 100% (95% CI 95.1% to 100%), and a positive predictive value of 97.1% (95% CI 93.3% to 99.3%).
Proportion of patients who met dyslipidemia case definition: N=773,081.
Chart abstraction vs algorithm results
Table 2 describes characteristics of patients who met the criteria for dyslipidemia and of those who did not. The data show that a greater proportion of seniors (≥65 years) have dyslipidemia (61.5%) compared with individuals between the ages of 40 and 64 (42.9%). Dyslipidemia was much more prevalent in males (56.7%) than in females (44.7%). A greater percentage of those living in rural areas than urban areas were found to have dyslipidemia (50.0% vs 45.2%, respectively). A large proportion of the population (43.3%) was classified as having obesity (BMI ≥30 kg/m2), with this group having a much higher prevalence of dyslipidemia at 64.9% compared with patients with a normal BMI (42.2%). A high prevalence of dyslipidemia was found in patients with hypertension (66.3%) and in patients with diabetes (57.5%). Patients with greater numbers of comorbid conditions also had higher dyslipidemia prevalence, with 67.9% of those with 4 or more comorbidities having dyslipidemia. A slightly higher prevalence of dyslipidemia was observed among patients in medium-sized practices (1000-1800 patients) compared with those in small (<1000) and large (>1800) practices, but these differences were not statistically significant.
Characteristics of patients with and without dyslipidemia and practice characteristics: Includes all patients who had at least 1 encounter at a participating practice in the 2-year study period.
The sensitivity analysis on the frequency of lipid testing by age, sex, and practice location found that males and females had similar testing rates between the ages of 50 to 75, with females aged 40 to 49 and 75 or older having slightly lower testing rates. Those in rural locations had a similar lipid testing frequency as patients in urban areas (details provided in supplementary materials, available from CFPlus*).
Approximately 97% of the LLAs prescribed to those with documented dyslipidemia were statins. The treatment status of patients with documented dyslipidemia, contrasted against the treatment status of patients who have other LLA-indicated conditions, is summarized in Table 3. Of those with documented dyslipidemia (excluding those with any other LLA-indicated condition) 32.5% were found to have been treated with an LLA. Of those with other LLA-indicated conditions (with or without dyslipidemia) 70.1% of those with diabetes and 40 or older were being treated with an LLA; approximately 69.2% of those who had had type 1 diabetes for at least 15 years were also treated. Of those individuals with CKD, CVD, and AAA, 63.2%, 71.8%, and 74.9% were treated, respectively. Finally, 61.0% of individuals with genetic dyslipidemia were found to have received pharmacotherapy.
Status of LLA treatment among patients with documented dyslipidemia and those with other LLA-indicated conditions
An FRS was calculated for each patient with documented dyslipidemia (Table 4). Among those stratified by low, intermediate, and high risk, the high-risk category had the highest proportion of individuals receiving treatment (65.0%). In the intermediate-risk group 48.7% were receiving treatment. The low-risk group had the largest number and highest proportion (77.2%) of untreated individuals. The sensitivity analysis evaluating whether classifying those with missing smoking status as non-smokers affected the proportions in each FRS category revealed that when patients with no smoking documentation are excluded there is very little change in the proportions in each category compared with classifying these patients as non-smokers (details provided in supplementary material, available from CFPlus*).
Status of LLA treatment among patients with documented dyslipidemia, according to Framingham risk score
Associations between the treatment status of patients with documented dyslipidemia and various demographic and clinical risk factors are summarized in Table 5. The fully adjusted model shows that the largest determinants of treatment for dyslipidemia are sex—males have a 1.95 times higher likelihood of being treated (95% CI 1.91 to 1.98) compared with females—and BMI, with those who have obesity having a statistically significantly increased likelihood of receiving treatment (adjusted odds ratio [AOR]=1.36, 95% CI 1.32 to 1.41). Individuals currently smoking were noted to have a 1.49 times higher likelihood of receiving treatment compared with individuals who have never smoked (95% CI 1.44 to 1.53). Those who were past smokers were also noted to have increased odds of receiving treatment (AOR=1.22, 95% CI 1.19 to 1.26). For each LLA-indicated condition, there were increased odds of receiving treatment for patients with these conditions compared with those without. The highest AOR was noted for those with diagnosed diabetes (AOR=4.51, 95% CI 4.44 to 4.66), and the lowest was observed in those with CKD (AOR=1.10, 95% CI 1.07 to 1.14). The number of comorbid conditions was only slightly predictive of being treated for dyslipidemia, with those having 2 or more comorbid conditions having at least a 1.10 times higher likelihood of being treated than those without a comorbid condition. Interestingly, those with 1 comorbid condition were less likely to have been treated (AOR=0.74, 95% CI 0.72 to 0.76) than those with no comorbid conditions. Patients seen at small- or medium-sized practices were only slightly more likely to have received treatment than patients seen at larger practices.
Associations between patients’ demographic characteristics, patients’ clinical characteristics, and practice characteristics and the odds of receiving LLA treatment among those with documented dyslipidemia: Includes all patients who had at least 1 encounter at a participating practice in the 2-year study period and had documented dyslipidemia.
DISCUSSION
National primary care data indicate that dyslipidemia is highly prevalent (50.0%) in Canadians aged 40 or older.6,22,23 Our estimates are comparable to a study by Joffres et al that used the Canadian Health Measures Survey (2007-2009) to assess the prevalence of dyslipidemia among Canadians using a similar definition consisting of a total cholesterol to HDL-C ratio ≥5, a measured LDL-C concentration ≥3.5 mmol/L, or use of lipid-modifying medications.6 They also noted a dyslipidemia prevalence of 45% among Canadians between the ages of 18 and 79 years. Given that our study excludes younger adults we assume a higher prevalence is to be expected.
In patients with documented dyslipidemia, we found that 42.8% were treated with an LLA. This is comparable to treatment rates reported in a study by Alabousi et al, which found that between 2007 and 2009 the treatment rate for Canadians with dyslipidemia was 40.9%.30 Unsurprisingly, when we investigated treatment status by FRS, we found that the high-risk category had the greatest proportion of individuals undergoing treatment (65.0%). Nevertheless, the remaining 35.0% of individuals in that risk category not receiving pharmacologic treatment represent a reasonable target for improvement. In the intermediate-risk group, 48.7% had been prescribed an LLA. The recommendations for treatment in this risk category are substantially more nuanced than for the high- or low-risk categories, as indicated in the updated 2021 Canadian dyslipidemia guidelines.4,12 Joffres et al noted that among individuals in the high-risk category, treatment was initiated in 49% of patients. Among those at intermediate risk in their study, only 20% were receiving treatment, even though approximately 80% of individuals met recommendations to receive treatment in that group.6 Hennessy et al also reported a discrepancy in those recommended for treatment in the intermediate category compared with those receiving treatment (67.7% and 14.9%, respectively).31 They found that among those in the high-risk category (for whom treatment is virtually always indicated), only 44.4% were undergoing treatment.31 While primary care providers should address therapy indications with those at high risk, our results are consistent with the literature indicating that patients with intermediate CVD risk may require additional consideration for treatment initiation.
Among patients with other LLA-indicated conditions examined in this study, a substantial proportion had not received LLA therapy from their primary care providers, ranging between 25.1% to 39.0% (Table 3). Previous studies have similarly reported on the undertreatment of various high-risk groups, primarily in those with diabetes.32 Our results highlight several opportunities for primary care providers to reduce CVD risk and improve outcomes in patients who are currently undertreated.
Risk of dyslipidemia (and of CVD) increases with age,8 and we see that reflected in these data, with the highest prevalence of dyslipidemia in those 65 or older. However, those between the ages of 40 and 64 composed the largest proportion of patients and therefore had the largest overall number of individuals meeting the dyslipidemia criteria. We saw a small but statistically significant increase in the odds of receiving treatment for each year increase in age (AOR=1.05, 95% CI 1.04 to 1.05). Joffres et al also noted a greater frequency of dyslipidemia treatment in those aged 60 to 79 years (48%) compared with those aged 40 to 59 years (18%).6 Although age is one of the variables used to stratify risk in the FRS calculation, the 2019 European Guidelines suggest that primary care providers should be mindful of hidden high relative risk in younger individuals in the context of low absolute risk.7
We found that more males than females met the case definition for dyslipidemia (56.7% vs 44.7%, respectively). The sensitivity analysis revealed there is slightly lower lipid testing frequency in females aged 40 to 49 and 75 or older when compared with males. As such, some of the sex difference observed in this study could be due to ascertainment bias. However, based on reports from comparable studies and the small magnitude of testing differences observed in the sensitivity analysis, we believe the estimate does reflect a true difference in dyslipidemia prevalence in males and females. Joffres et al found similar results, with a 43% prevalence in males and a 29% prevalence in females.6 In the present study, for those with dyslipidemia there was a 1.95 times higher likelihood of males receiving LLA treatment than females. In 2010 Goodman et al noted gender disparities among Canadians in dyslipidemia treatment target achievement, with women being less likely to meet treatment targets than men, despite adjustment for age and confounding variables.16 Our results indicate that this disparity appears to be persistent in Canada.
Internationally and in Canada, while CVD causes more deaths in women than in men, men are more likely to die of CVD before the age of 65 years.6,7,19,33 Dyslipidemia and cardiovascular risk for women are complicated by menopause, which can be associated with a substantial rise in LDL-C levels and decreased HDL-C levels.8,33 The risk of CVD is deferred to a later age due to the various physiologic shifts postmenopause.7,8,33 Primary care providers need to review cardiovascular risk factors, including dyslipidemia, in all postmenopausal women and be mindful not to miss patients in early menopause or with premature ovarian insufficiency. Similar studies using administrative data have indicated that higher clinic volume may improve guideline-concordant care for patients at risk of CVD.33 Our analysis found that patients cared for by providers with small- to medium-sized practices are slightly more likely (1.1 and 1.07 times, respectively) to receive an LLA compared with patients receiving care in larger practices. More research is needed to explore practice-level factors that may contribute to the prevention and management of CVD with LLAs.
Limitations
This study has several limitations. The nature of the case definition and various criteria to investigate comorbid conditions do not provide perfect accuracy. There may be some misclassification in diseases and treatment status of patients. In addition, some CVD risk factors used to derive the FRS calculation may have gaps in data quality, leading to lower risk scores owing to missing data (eg, smoking or diabetes status). Therefore, there is likely an overestimation of the number of patients in the low-risk category. These variables are likely not missing at random but reports that describe these data quality gaps are currently unavailable.
Due to resource limitations the validation work was completed using data from 1 interprofessional primary care practice with 25 primary care providers in Ontario, Canada. As such, the robustness of the case definition may differ in clinics from other jurisdictions or using different EMR software. In addition, the expert panel had only 2 members rather the recommended 3. However, there were few disagreements in the chart review and consensus was easily reached.
Our data represent what is entered in primary care EMRs and may not reflect all patient care. For instance, there may be relevant information in paper form, or the patient may receive treatment from specialists (eg, cardiologist, endocrinologist). We also did not capture any information related to nonpharmacologic management or shared decision making in cases where patients may have been offered but declined or discontinued LLAs based on their values or appropriate clinical circumstances. It is entirely possible that many individuals who are not currently being treated were offered treatment in the past. We also did not assess whether those taking lipid-lowering therapy were attaining treatment targets, which is a somewhat controversial area in family medicine given conflicting recommendations.4,12 More than half of the study population did not have full postal codes as part of their data. As such, we were not able to derive a geographic measure of social and material vulnerability and could not describe how socioeconomic status influenced the prevalence and treatment of dyslipidemia.
Despite these limitations, our results have several strengths by providing helpful information and useful improvement targets for primary care providers. In particular, the large cohort and rich data set allow smaller teams or regional networks to use these findings as a benchmark for their data to prioritize and measure their improvement strategies. Our validated case definition can also be used to support other ongoing research on comorbidities and chronic disease management in primary care settings.
Conclusion
The study provides an updated look at the prevalence and treatment of dyslipidemia among Canadians. Half of patients, aged 40 years and over, have dyslipidemia, with an even higher prevalence observed among older adults, males, and those with obesity or other chronic conditions. There remain substantial treatment gaps among those with documented dyslipidemia, principally in those with intermediate and high FRS calculations. Particular attention should also be paid to those at higher risk for not receiving treatment, including female patients and those within normal BMI ranges.
Footnotes
↵* Supplementary material is available from https://www.cfp.ca. Go to the full text of the article online and click on the CFPlus tab.
Contributors
All authors contributed to conceptualizing and designing the study; to collecting, analyzing, and interpreting the data; and to preparing the manuscript for submission.
Competing interests
Dr Alexander G. Singer works as a knowledge translation consultant with the Canadian Cardiovascular Society.
This article has been peer reviewed.
Cet article a fait l’objet d’une révision par des pairs.
- Copyright © 2024 the College of Family Physicians of Canada