Dyslipidemias and 10-Years Cardiovascular Risk Scores in Adults in Asmara, Eritrea: Findings From a Community-Based Cross-Sectional Study

Background : The objective of this study was to estimate the prevalence of dyslipidemias and associated 26 factors in adults ( ≥ 35 to ≤ 85 years) living in Asmara, Eritrea. 27 Methods : A total of 384 (144 (%) males and 242 (%) females, mean age ± SD, 68.06±6.16 years) 28 respondents were randomly selected after stratified multistage sampling. The WHO NCD STEPS 29 instrument version 3.1 questionnaire was used to collect data. Measurements/or analysis including 30 anthropometric, lipid panel, fasting plasma glucose (FPG), and blood pressure (BP) were also 31 undertaken. 32 Results : The frequency of dyslipidemia in this population was disproportionately high (87.4%) with the 33 worst affected subgroup in the 51-60 age band. The level of awareness was also low. In terms of 34 individual lipid markers, the proportion were as follows: HDL-C (40 mg/dL men and 50 mg/dL females) 35 (55.2%); TC ≥ 200 mg/d (49.7%); LDL ≥130 mg/dL (44.8%) ; TG≥150 mg/dL (38.1%). The mean ± SD, 36 for HDL-C, TC, LDL-C, non-HDL-C, and TG were 45.28±9.60; 205.24±45.77; 130.77±36.15; 37 160.22±42.09 and 144.5±61.26 mg/dl, respectively. Regarding NCEP ATP III risk criteria, 17.6%, 38 19.4%, 16.3%, 19.7%, and 54.7% were in high or very high-risk categories for TC, Non-HDL-C, TG, 39 LDL-C, and HDL-C, respectively. Among all respondents, 59.6% had mixed dyslipidemias with 40 TC+TG+LDL-C dominating. In addition, 27.3%, 28.04%, 23.0%, and 8.6% had abnormalities in 1, 2, 3 41 and 4 lipid abnormalities, DBP (aOR 1.04 mmHg (1.00-1.09)=0.001) and increasing FPG (aOR 1.02 1 mg/dL, 95% 1.05, p=0.001). 49 Conclusion: High frequency of poor lipid health may be a prominent contributor to the high burden of 50 CVDs – related mortality and morbidity in Asmara, Eritrea. Consequently, efforts directed at early 51 detection, and evidence-based interventions are warranted. 52 57


Introduction
factors for Ischemic Heart Disease (IHD) in SSA 9 . Mechanistically, dyslipidemia triggers endothelial 90 dysfunctiona process that is central to the pathogenesis of atherosclerosis, thrombosis, and 91 hypertension. 92 Despite the contribution of dyslipidemia to the high and rising burden of CVD in SSA 4,6 ; the condition is 93 under-diagnosed, under-treated and under-described. This represents a missed opportunity given the fact 94 that targeting risk drivers of CVD at the population level by a combination of simple, low cost, efforts 95 could avert more than 50% of the associated morbidity and mortality 10 . This, therefore, cannot be 96 overemphasized: poor lipid health in adult populations in SSA is now a true epidemic and public health 97 crisis that both patients and clinicians must face. Previously identified barriers to addressing this problem 98 include lack of awareness (among the public and health-care professionals); high cost of diagnosis and 99 treatment; a dearth of well-trained health personnel; lack of local clinical practice guidelines; under-100 treatment and a limited understanding of its epidemiology 4 . Overall, the lack of reliable health 101 information/statistics and a severe lack of community-based epidemiological data is a source of serious 102 concern as it handicaps public health strategies directed at prevention and treatment/or management of 103 dyslipidemia in the region. 104 Much of the description in the foregoing paragraph applies to Eritrea. At present, very little is known 105 about the risk factors associated with CVD in any population in Eritrea. The lack of data is extremely 106 concerning given the fact that World Health Organization (WHO) fact-sheets have consistently shown 107 that CVD-related mortality in the country is disproportionately high (388.1 vs. 282.2 per 100 000 in 108 males and females, respectively) 11 . Interestingly, the country has one of the lowest prevalence of 109 overweight/obesity (mean BMI = 20.5 (95% CI: 19.9-21.1 and BMI ≥ 25 kg/m 2 = 17.7% (95% CI: 14.7% 110 -20.2%) in SSA 11,12 . The prevalence of other known drivers such as tobacco use, irresponsible alcohol 111 consumption, hypertension, and Diabetes mellitus are modest or low 3 . One facet of this problem that has 112 received little attention is the burden of dyslipidemia and its possible contribution to the excess burden of 113 CVD in the country. 114 Therefore, this study was designed to generate population-level data on the burden of dyslipidemia in the 115 adult population in Asmara, Eritrea. Beyond the focus on dyslipidemia, using multivariable risk 116 scores/prediction algorithms to identify persons at higher risk is a well-established intervention strategy 117 and has proved to be cost-effective in multiple jurisdictions 13,14 . Therefore, we computed Framingham 118 CHD risk scores (FCRS) of the participants. Data on these markers can be crucial in designing evidence-119 based, context-specific community-level and/or individualized interventions. The information can also be 120 leveraged in the future to evaluate whether CVD risk is declining, stagnating, or even increasing. Further, 121 estimating CVD risk using FCRS presupposes the collection of data on an expanded list of risk factors. 122 To this end, data regarding the prevalence and distribution of individual risk factors such as obesity, 123 fasting blood sugar (FPG), blood pressure (BP), among others, are also presented.  Sample size determination and sampling procedure 133 "A single population-proportion formula was used to determine the sample size 15 ." Using a proportion of 134 50% (dyslipidemia) (we used 50% prevalence because there were no credible data on the prevalence of 135 dyslipidemia in Eritreathis resulted in the highest sample size), 95% confidence interval (CI), 5% type 136 I error level, 80% power and adjustment for response rate; the sample size was calculated to be 384 137 participants. To identify eligible study participants, a stratified multistage sampling design was followed.

138
Briefly, the total sample size was proportionally allocated to the 13 sub-zones. After eliminating sparsely 139 populated Zip-codes within each Sub-zone, one Zip code (enumeration area (EA)) was selected using a 140 simple random sampling technique (computer-based random number generator Japan). After 10 minutes of rest, 3 measurements were taken after an interval of ~5 minutes. The    Prediabetes (FPG ≥100 and ≤125 mg/dL) (<5.6 -≤ 6.9 mmol/L); undiagnosed DM (FPG ≥125 mg/dL) 206 (≥7 mmol/L). Taking into account the cost of the study and the level of sensitivity and specificity 207 required for presumptive diagnosis of DM, HbA1c analysis was restricted to FPG ≥ 125 mg/dL. As per 208 ADA criteria, HbA1C < 5.7% was regarded as normal; HbA1C between 5.7% -6.4% was classified as pre-209 diabetes and HbA1C > 6.5% was classified as undiagnosed DM. CVD risk markers including HDL, age range, hypertension treatment, smoking, and TC. Similar, to 214 Reiger et al., Framingham CVD Risk Score was ascribed to participants based upon the low risk (< 3%); 215 moderate (≤3% to <15%); high (≥15 to <30%), and very high (≥30%) 19 .   Table 1 for additional information.

252
The relationship between specific demographic characteristics, anthropometry, clinical measurements, 253 and specific lipid markers was also evaluated. compared to respondents with normal FPG; lower HDL-C, and higher TC/HDL ratio. Finally, increasing 259 age (to ≤ 60 years) was characterized by higher values in TC, LDL-C, non-HDL-C, and TC/HDL ratio.

260
See Table 2 for additional information.  As seen in Table 4, mixed dyslipidemia, defined as the presence of ≥ 2 lipid abnormalities, was also 277 analyzed. Among all respondents, 59.6% (95% CI 54.6% -64.6%) had mixed dyslipidemias. Most  Factors associated with elevated Non-HDL-C, TG, and TC 284 We summarize here the results of the multivariate models in  Table 6 for additional information.

Framingham Risk Scores: Magnitude and Relationships
Framingham risk scores for all the respondents were computed as highlighted in the previous section. In the overall sample (N = 384), 166(43%) of the respondents were in the low-risk strata; 160(41.5%) in the moderate risk strata; 49 (12.7%) in the high-risk strata and 11(2.8%) were in the very high-risk strata. Figure 2A. Disagregation of the data for selected lipid abnormalities is shown in Figure 2B. Briefly, 23.1% of the respondents with elevated LDL and 21.4% of patients with high TC/HDL ratio were in the high-risk category.
Socioeconomic factors may play differential underpinning roles in CVD. Consistent with the data presented above, a large proportion of patients with No formal education or divorced/widowed had a high 10-year risk of CVD -27.3% and 30.8% respectively. As expected, a connection between Age, FPG, and FRS was also demonstrated. See additional information in Table 7.

Discussion
Although more than 80% of the global burden of CVD in LMIC, knowledge of the importance of risk factors is largely based on extrapolations from high-income countries (HIC) 4 27 . In other words, the frequency of dyslipidemia in some parts of Africa is similar or even higher than what has been reported in some high-income countries (HIC) -(33-75%) 28 . These estimates should raise concern considering the well-documented association between dyslipidemia and CVD. Overall, but with notable exceptions 22,30 , HDL-C was the most common isolated lipid abnormality 8 . Emphasizing this point, some investigators have suggested that over the last 50 years, a decline in HDL-C concentration has been gradually occurring in Africa 31 . Separately, the Heart of Soweto Registry Study reported that low HDL-C is a common presentation in individuals with non-communicable heart disease 8 . Although causality was not inferred, and extrapolation of the data to this setting may be dubious; the research raised pertinent issues which should be explored in this setting.
Although the high frequency of low-HDL in populations in SSA is well documented; the importance of this phenomenon is still elusive. Part of the problem is attributable to the ambiguity concerning the relationship between HDL-C and CVD. For example, genetic studies using Mendelian randomization and longitudinal studies present contrasting results 32 .
Nevertheless, and considering the role of HDL-C in reverse TC transport, its anti-oxidant, antithrombotic, and anti-inflammatory properties; its cardioprotective role is presumed to be substantial. At present, research implicates a strong genetic contribution (~50%) to the observed variability of serum HDL-C at the population level 33 . Presumably, 50% of the observed variation has been linked to a diverse range of modifiable risk factors including insulin resistance/type 2 diabetes mellitus (T2DM), BMI > 25 Kg/m 2 , sedentarism/physical inactivity, βblockers, progestational agents, cigarette smoking, and very high carbohydrate intakes 34 . Other inputs include infection and inflammation 35 .
In our study, we established a positive association between low HDL-C, elevated FPG, and employment status (predominance in the unemployed portion of the population). Another notable outcome was the weak association between low HDL-C and BMI categories. Similar to other studies 34,36 , consumption of alcohol was associated with increased concentration of HDL. In general, the mechanisms underpinning the positive correlation between alcohol consumption and HDL-C concentrations are not known, although it has been hypothesized that alcohol may ApoB/apoA1 ratios provide equivalent information about CVD risk. Further, the evidence that LDL -C prospectively predicts hard CVD events (coronary death, MI, and stroke) is unequivocal. However, whether or not TG levels are a causal risk factor for CVD remains unclear. Regardless, we have to note that TG, TC, and LDL-C existed, rarely, in isolation in this setting. Therefore, the high proportions and, to some extent, mean values of TC, LDL-C, TG, TC/HDL-C, and non-HDL-C observed in this setting should raise concern. To this end, dissecting the factors associated with abnormalities in relevant lipid parameters can provide useful data for targeted public health intervention.
To a large extent, most associations of TG, TC, TC/HDL-C, LDL-C, and Non-HDL-C were in the expected direction. High TC was independently associated with alcohol consumption, hypertension, and FPG. Similar to other studies 39 , elevated LDL-C was associated with alcohol consumption, WC, and hypertension. These risk factors were also associated with non-HDL-C and TG (add sex) in this study. Remarkably, a large proportion of participants with elevated LDL-C were in the Framingham risk scores high-risk category. Another interesting relationship was the observed association between TC/HDL-C ≥ 5 ratio and sex (higher in males); WC, hypertension, and FPG. Despite the broad agreement between this study and other studies, notable exceptions were observed. For example, BMI ≥25 kg/m 2 had only one association (LDL-C). This was in contrast to studies that have uncovered a significant relationship between elevated BMI, high TG, and low HDL-C 40 . We are unable to provide definitive explanations why BMI is a poor marker of dyslipidemias in this setting. However, we found a significant relationship between WC and multiple dyslipidemias in the multivariate analysis -Non-HDL-C, LDL-C, TC/HDL, and TG. As previously noted 41 , the use of WC for public health screening or clinical evaluation of patients is still limited in Eritrea. In this regard, the current study merely adds to the evidence on its utility within in setting.
Furthermore, troubling associations and patterns were apparent in this population. The high frequency of dyslipidemia in women (high regardless of menopausal status), the unemployed, those who are divorced/or widowed, and those with lower levels of education points to a gathering problem among a vulnerable subgroup of the population. The same pattern was observed when FRS was computed. The clustering of CVD risk factors among the unemployed, in populations of low socioeconomic status, or among the less educated strata of the society is well documented 34 . According to some authors, education mediates the risk of CVD through urbanization, unemployment, access to information/awareness, food, social support and cohesion, and individual health behaviors. In any case, our data support the possibility that physical inactivity/sedentarism and excess weight gain (particularly abdominal obesity) are significant contributors to dyslipidemia in this setting.
Another important finding was the fact that combined dyslipidemia is more common in this population (68.6%) compared to single lipid abnormalities. For example, 28.04%, 23%, and 8.6% of the study participants had abnormalities in two, three, and four lipid components, respectively. The most common combination was high TC + TG + LDL-C (20.9%). The proportion of participants with high TC+ low HDL-C and high TG + low HDL-C was also  20,44 . Unfortunately, we could not locate comparable community-based studies on the epidemiology of mixed dyslipidemias from the region. Therefore, rigorous population-based prospective investigations are necessary to determine the risk associated with the co-presence of specific lipid abnormalities.

Strengths and Limitations
This study is not without limitations. First, the cross-sectional nature of the study limits the dissection of causality. In addition, the fact that the population was mostly composed of urban residents limits the generalizability of our findings. The use of a researcher-administered questionnaire to capture data on specific variables may be affected by the recall, social desirability, and outcome misclassification biases. Lastly, the Framingham risk score and Friedewald equation for LDL-C estimation have not been validated in this population; hence the results should be used with caution. Despite the above limitations, and in the absence of longitudinal studies, this investigation represents a major first step towards getting baseline on lipid profiles in Asmara, Eritrea. Attempt to analyze the frequency of mixed dyslipidemia along with Framingham risk score also adds to its uniqueness.

Conclusion
This study uncovered many important findings. First, the frequency of dyslipidemia in this population was disproportionately high (87.4%). In terms of relative frequencies, the most prevalent lipid abnormality was low HDL-C (55.2%) followed by high TC (49.7%); high LDL

Additional Information
The corresponding author is responsible for submitting a competing interests statement on behalf of all authors of the paper. The authors have no conflict of interest to declare on this study.
Material support was obtained from the Eritrean Ministry of Health.

Supporting Documents
The dataset supporting the conclusions of this article are available from the corresponding author on reasonable request.

Ethics approval and consent to participate
Ethical approval for the study and experimental protocols used was obtained from Eritrean Ministry of Health (MOH) research ethical committee. Informed consent was obtained from all participants. During the study, strict adherence to approved laboratory protocols was observed.