Prevalence and factors associated with lipid-lowering medications use for primary and secondary prevention of cardiovascular diseases among Malaysians: the REDISCOVER study

The REDISCOVER (Responding to Increasing Cardiovascular Disease Prevalence) is an ongoing epidemiological prospective study with the duration of 15 years. The data was collected every 3 years, with the baseline collections from 2007 to 2011. This article presents the cross-sectional analysis of the primary data of RESDISCOVER study, which included sample of 11,288 respondents, aged 30 years and older from the baseline recruitments.

The REDISCOVER study involves 22 rural and 18 urban communities from five states across West and East Malaysia. The states representing West Malaysia were Selangor, Negri Sembilan, Pahang and Kelantan, while Sabah represented East Malaysia. These states were chosen to ensure adequate representation of the major ethnic groups in Malaysia. Major ethnic groups in West Malaysia, also known as Peninsular Malaysia are Malays, Indians and Chinese. Three states of West Malaysia (Selangor, Negri Sembilan and Pahang) have an adequate mixture of Malays, Chinese and Indians, while Kelantan population is predominantly Malay.

The major ethnic groups in Sabah (East Malaysia) were Kadazan-Dusun, Bajau, and Murut. These groups, along with other ethnic minorities in Sabah, are known as indigenous people. The sampling frame for this study were Malaysians aged 30 and above from the 40 communities.

The participants were recruited in a standardized four-staged process; selecting the states, then communities, followed by the households within the communities, and then the individuals within the households. In each selected state and community, announcements and invitations were made through local community leaders. All household members aged 30 years and older received written invitations to attend screening sessions at local community centres. They were requested to fast for 8 hours in preparation for fasting blood sampling. The participants were screened for eligibility at the community centres, and written informed consent was obtained.

All investigators and interviewers were trained on the study procedures. Researchers utilised standardised data collection forms that included demographic characteristics such as age, gender and educational attainment, as well as clinical details such as smoking status, history of diabetes mellitus, hypertension, CVD, and the use of LLM.

For anthropometric measurements, waist and hip circumferences were measured using a non-stretchable measuring tape while the participants stood in a relaxed position, with arms on the side. The measurements were recorded to the nearest 0.1 cm. Automatic digital blood pressure (BP) monitors (Omron HEM-757) were used to measure blood pressure after a five-minute rest. The BP on the right arm was measured twice, 2 minutes apart, with the participants in a seated position. The mean of the two measurements was taken as the BP reading for the participant.

Fasting venous blood samples were taken for lipid profiles [total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-c), low-density lipoprotein cholesterol (LDL-c)] and plasma glucose level. The samples were centrifuged within 2 hours of collection, stored at the study sites in the minus 20-degree Celsius freezers, then transported frozen to the main laboratory at the Faculty of Medicine, Universiti Teknologi MARA, Malaysia. The samples were analysed using an automated clinical chemical analyzer (Cobas Integra 400 plus, Roche Diagnostic, Basel, Switzerland). The LDL-c was calculated using the Friedewald equation (for TG ≤ 4.5 mmol/L) [16].

Educational attainment was classified into four groups: no formal education, primary, secondary and tertiary. Primary education was defined as schooling between aged seven and 12 years old, while secondary education was schooling from 13 to 17 years old. Tertiary education was defined as the attainment of college or university qualification. Communities with a population of ≥10,000 were defined as urban, while those with populations of < 10,000 as rural [17]. Household income was classified according to the Department of Statistics Malaysia, where the income groups were classified into three categories: the bottom 40% (household earnings of Ringgit Malaysia (RM) ≤ 4849), medium 40% (household earnings of RM 4850–10,959) and top 20% (household earnings of RM ≥ 10,960) [18].

As for the clinical characteristics, CVD were defined as self-reported ischaemic heart disease and/or stroke. Smoking status was categorized as current smoker (had smoked any tobacco product in the last 5 years), non-smokers (had never smoked) and ex-smokers (had stopped smoking for more than 5 years). Body mass index (BMI) was classified as the following: underweight < 18.5 kg/m², normal 18.5–22.9 kg/m², overweight 23–27.4 kg/m², or obese ≥27.5 kg/m² [19, 20]. Abdominal obesity was defined as waist-hip ratio (WHR) of ≥0.90 for males and ≥ 0.85 for females [21].

The abnormal fasting lipid profile was defined according to the Malaysian Clinical Practice Guideline, which were: hypercholesterolaemia (TC > 5.2 mmol/L); hypertriglyceridaemia (TG > 1.7 mmol/L) and low HDL-c (HDL-c < 1 in males; < 1.2 mmol/L in females) [3]. This guideline also defines elevated LDL-c according to CVD risk profiles, where LLM is indicated when LDL-c > 1.8 mmol/L for very high-risk category, > 2.6 mmol/L for high-risk category and > 3.4 mmol/L for intermediate risk group [3]. For low-risk category, the guideline did not specify the LDL-c cut-off point for LLM, and recommended LLM therapy based on joint decision making between prescribers and patients; taking into account the risk and benefits of LLM therapy [3]. For the purpose of this study, the highest level, i.e., LDL-c > 3.4 mmol/L was chosen to define “elevated LDL-c”; as above this threshold, LLM is indicated for patients from intermediate, high risk, and very-high risk categories. Hypertension was defined as mean systolic blood pressure (BP) of ≥140 and/or mean diastolic BP ≥90 mmHg and/or self-reported hypertension and/or self-reported history of taking antihypertensive medications in the last month.

The participants were classified as having diabetes if they fulfilled any of the following criteria: fasting plasma glucose of ≥7.0 mmol/L and/or self-reported diabetes and/or taking any diabetes medications for the past 1 month. The Global CVD risk of participants without existing CVD was calculated using the 10-year FRS-CVD score [7]. Classifications were as follows: low risk (10-year FRS-CVD score <  10%), intermediate risk (10-year FRS-CVD score 10–20%) or high risk (10-year FRS-CVD score >  20%). Based on the Malaysian Clinical Practice Guidelines on Management of Dyslipidaemia, individuals with diabetes or a 10-year FRS-CVD score > 20% are classified as high risk, while those with existing CVD are automatically classified as very high-risk individuals [3]. For the purpose of this study, the three groups (diabetes, FRS-CVD score > 20% and existing CVD) will be collectively referred as high-risk individuals. Local guideline recommends LLM use at the outset among these high-risk individuals [3].

LLM use was defined as self-reported use of either one or any combinations of lipid lowering medications, such as statin, ezetimibe, or fibrate.

The STATA software version 14 was used for statistical analysis [22]. The cases without specific variable data (missing data) were excluded from analysis for that variable only, using pairwise deletion. Descriptive statistics were presented as frequencies and percentages for categorical variables. For numerical variable, mean ± standard deviation (SD) was used for normally distributed data, or median with interquartile range (IQR) for non-normally distributed data, respectively. Chi-square test or Fisher’s exact test was used to test the difference in proportions. Simple and multiple logistic regressions were used to estimate the crude and adjusted odds ratio (OR) for the factors associated with the use of LLM. A p-value of less than 0.05 was considered statistically significant.

Approximately 19,000 invitations were sent to household members above the age of 30 years to attend screening sessions at their local community centres. The response rates were between 60 and 70%. A total of 11,288 participants were included in the analysis. Their sociodemographic and clinical characteristics are presented in Table 1. The median age (IQR) was 52 (15) years. Majority of the participants were female (56.2%), and of Malay (72.5%) ethnicity. There was an almost equal distribution of urban (51.9%) and rural (48.1%) participants. Most of the participants (89.2%) were from the bottom 40% household income group. As for the clinical characteristics, 5.3% of participants had existing CVD. Among those without CVD, 23.4% were in the FRS-CVD high risk category.

Table 2 presents the use of LLM among high-risk participants. In each group, the majority of participants were not on LLM. Out of those with existing CVD, 74.2% were not on LLM for secondary prevention of CVD. As for participants with diabetes, only 17.1% were on LLM. Among those with high FRS-CVD risk score, 89.5% of participants were not on LLM.

The factors associated with LLM use among all participants are shown in Table 3. Those who received secondary education (OR = 2.46, 95% CI: 1.29–4.70) and tertiary education (OR = 2.25, 95% CI: 1.06–4.78) were more likely to be on LLM, compared to those with no formal education. Participants residing in rural areas had less odds of being on LLM (OR = 0.58, 95% CI: 0.41–0.82) compared to urbanites.

Participants who had CVD risk factors, such as obesity (OR = 1.80, 95% CI: 1.15–2.83), diabetes (OR = 2.38, 95% CI: 1.78–3.19), and hypertension (OR = 2.87, 95% CI: 2.09–3.95) were more likely to be on LLM. As for secondary prevention, those with existing CVD had increased odds of being on LLM (OR = 3.30, 95% CI: 2.26–4.84). Participants with elevated LDL-c (OR = 0.66, 95% CI: 0.44–0.98) and low levels of HDL-c (OR = 0.72, 95% CI: 0.53–0.99) were less likely to be on LLM (OR = 0.66, 95% CI: 0.44–0.98).

The FRS-CVD score was calculated for the primary prevention group, and included in a separate multiple logistic regression model. Table 4 presents factors associated with LLM use among the primary prevention group. After adjusting for socio-demographic confounders, participants in the high-risk category had higher odds of being on LLM (OR = 3.81, 95% CI: 2.78–5.23), compared to those in the low-risk category. Similar to the model that included the overall cohort (primary and secondary prevention groups), this model (primary prevention group only) found that education level and locality (urban versus rural) were significant determinants of LLM prescription. In addition, income group was also a significant factor in this model (primary prevention group), where those who earned equal or more than RM 4850 were 1.5 times more likely to be on LLM, compared to those from the lower income group (OR = 1.54, 95% CI: 1.06–2.24).

The main strength of this study is the large sample size with good response rate (60–70%) from various parts of East and West Malaysia, which provided good representation of the population. Although the response rate is acceptable, recruitment via voluntary participation may introduce non-response bias. This is likely minimal as the characteristics of our study population were comparable to another Malaysian study with similar setting, where 52.4% of their study population were females, and 42.3% were rural residents [15], whereas our study comprised of 56.2% females and 48.1% rural residents. While another large local population study used non-fasting, finger-prick capillary blood lipids [15], this study demonstrated robustness of laboratory measurements by using fasting venous blood sampling as recommended by the Malaysian guideline [3], which is commendable for a large epidemiological study.

This study had several limitations. The participants were recruited from health screening sessions at the local community centres. Self-reporting was used to collect several variables, like LLM use, which may have led to recall bias. Although participants were reminded to bring their medications to these sessions for verification by the interviewers, the vast majority did not. Interviewers were trained to ask specific questions to clarify the information provided by the participants, such as the reason for their LLM use, to minimise recall bias. Data on contraindications to LLM were not available in this study. Therefore, those high-risk individuals who were not on LLM may be contraindicated to receive LLM.

The participants who attended these voluntary community health screening events may have a greater awareness of their health which may also help to minimise recall bias. This may also explain the findings of our study which had lower prevalence of smokers (11.5%) compared to the results from the Malaysian National Health and Morbidity Survey (21.3%) [25]. Another limitation includes overrepresentation of Malays in this study. Malays, Chinese and Indians represent 63.1, 24.6 and 7.3% of Malaysia’s population, respectively [17], while this study comprised 72.5% of Malay participants. The use of survey weight could be employed to overcome this limitation, nevertheless it was not done in this study. Finally, the cross-sectional analysis of this study would only reflect the association between participant characteristics and LLM use, not the causal relationship. Thus, the results should be interpreted in this context.

The REDISCOVER study provided an insight to healthcare providers and policy makers regarding the widespread underutilisation of LLM among Malaysian adults, especially those who are in the high-risk categories. Urgent public health measures are required to address this issue. Two main possible reasons for the underutilisation could be due to prescriber factors or individual patient factors. A Malaysian study showed that while the majority of clinicians used dyslipidaemia guidelines, there were wide variation in their practice with some clinicians did not prescribe based on the guidelines [41]. Furthermore, another study reported a high proportion of therapeutic inertia for LLM (61%), which resulted in lower odds of achieving LDL-c target [42]. Various factors have been proposed to contribute to treatment inertia, including physician’s fear of side effects, as well as the availability of appropriate medications [42]. Inappropriate prescriptions were also observed in a previous study in the United Kingdom where most patients were initiated on statins without global CVD risk assessments, and some overtreatment also occurred in the low-risk patients [43]. A Malaysian study reported better results where 79.7% of clinicians performed risk-scoring prior to statins prescription [41]. In order to resolve the possible knowledge gap that could have led to inappropriate LLM use, the results of this study should be conveyed to prescribers via continuous medical education programmes. With regards to the individual patient factors, the findings of this study on sociodemographic disparities provide policymakers with an insight to ensure equitable access to LLM for individuals with lesser formal education, low-income earners and rural residents. In terms of future research, certain variables that could influence the use of LLM, such as medication adherence, health belief towards LLM and health literacy should also be included, to help identify patient factors contributing to suboptimal LLM use. Future research should also include collaboration with the participants’ health care providers to verify the type of LLM prescribed which will provide more credence to the study. Factors that influence prescribing of LLM among health care providers, such as contraindication to LLM, their knowledge on guideline recommendations and personal belief towards LLM should also be explored.