Hypertriglyceridemia and Other Plasma Lipid Profile Abnormalities among People Living with Diabetes Mellitus in Ethiopia: A Systematic Review and Meta-Analysis

Background Dyslipidemia is one of the leading causes of cardiovascular complications in diabetes mellitus (DM) patients. Though it is a major public health problem in Ethiopia, there is no a nation-wide study to determine dyslipidemia among DM patients yet. Therefore, this systematic review and meta-analysis intended to estimate the prevalence of hypertriglyceridemia and other plasma lipid abnormalities among people living with DM in Ethiopia. Methods We systematically searched PubMed, Google Scholar, African Journals Online, Hinari, and direct Google. Studies conducted until May 9, 2020, that reports the prevalence of dyslipidemia among people living with DM were included. The DerSimonian and Laird random-effects model was used to determine the pooled prevalence of lipid profile abnormalities. Heterogeneity was checked using the I2 statistic, whereas publication bias was tested by funnel plot and Egger's test. Besides, subgroup and sensitivity analyses were performed. Results We used 18 primary studies, including 4961 participants living with DM, which met the eligibility criteria for the meta-analysis of hypertriglyceridemia. The estimate of hypertriglyceridemia (≥150 mg/dl) was 48.15% (95% CI: 38.15-58.15, I2 = 98.4%) after performing the main meta-analysis using the random-effects model. The subgroup analysis showed a higher pooled estimate of hypertriglyceridemia among T2DM (57.80% (95% CI: 50.50-65.10), I2 = 92.5%), studies that used probability sampling technique (59.09% (95% CI: 43.58-74.59), I2 = 98.6%, p < 0.001), and studies from primary data sources (51.43% (95% CI: 40.72-62.13), I2 = 98.0%, p < 0.001). Moreover, the estimated pooled prevalence of the total plasma cholesterol (TC ≥ 200 mg/dl) was 34.08% (95% CI: 28.41-39.75, I2 = 92.4%), LDL − C ≥ 100 mg/dl was 41.13% (95% CI: 27.15-55.11, I2 = 98.8%), and HDL ≤ 40 mg/dl for men and ≤ 50 mg/dl for women was 44.36% (95% CI: 31.82-56.90, I2 = 98.8%). Conclusions The pooled prevalence of hypertriglyceridemia and other lipid abnormalities among DM patients was relatively high in Ethiopia. It strongly suggests the need to give maximal attention to the adherence of DM management to reduce the circulatory lipid profile abnormalities and subsequent complications. Prospero Registration. CRD42020182291.


Background
Dyslipidemia is a derangement of lipoprotein metabolism characterized by an increased level of serum triglycerides (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), and decreased level of high-density lipoprotein cholesterol (HDL-C) [1]. It is a risk factor for cardiovascular diseases (CVDs) [2,3]. Acquired dyslipidemia is caused by sedentary lifestyle such as an unhealthy diet and chronic illnesses such as obesity, chronic kidney disease, hepatic diseases, and DM [4][5][6]. In return, dyslipidemia contributes to the occurrence of CVDs in DM patients [7].
Dyslipidemia in DM patients is largely attributed to both increased production and delayed catabolism of very-lowdensity lipoprotein (VLDL) secondary to resistance and a relative deficiency of insulin, a key hormone that regulates lipid metabolism [3,8]. Physiologically, insulin reduces VLDL production by inhibiting the activation of hormone-sensitive lipase at adipose tissue and decreases the circulating nonesterified free fatty acids from adipocytes. Thus, the production of nascent VLDL and apoprotein B100 (Apo B100) is inhibited because of limited free fatty acids in the circulation [9]. Increased hepatic lipogenesis also increases liver cholesterol [10]. Besides, insulin activates lipoprotein lipase (LPL) which metabolizes circulatory triglyceride and VLDL to free fatty acid and glycerol and also increases the expression of LDL receptors to facilitate the uptake of VLDL remnants by the tissues [3]. In DM patients, low, absence, or insulin resistance results in a rise of triglyceride and VLDL. The normal inhibitory effect of insulin on hepatic Apo B100 production and triglyceride secretion is lost, and the secreted VLDL becomes higher than utilization [11]. Insulin resistance causes loss of potent stimulatory effect of insulin on LPL and contributes to the increment of triglyceride in the plasma [12,13]. Stimulation of LDL clearance via increasing LDL-C receptor expression and activity by insulin is also lost because of insulin resistance [3].
Diabetes mellitus is associated with vascular problems such as peripheral arterial disease [14] and hypertension [15]. Hypertriglyceridemia combined with low HDL-C or high LDL-C in DM populations contributes to life-threatening vascular complications [8,16,17] including ischemic stroke [18,19]. Besides the morbidity and mortality, hypertriglyceridemia increases the health care costs of people living with DM [20].
Globally, dyslipidemia causes around 2.6 million deaths and 29.7 million disability-adjusted life years (DALYs) [24,25]. The prevalence of dyslipidemia, among DM patients, is 67.1% in China [26], 86% in Bangladesh [27], 88.9% in Thailand [28], and 89% in India [29] whereas, in Ethiopia, hypertriglyceridemia ranges from 14.2% [30] to 72.5% [31]. Apt management is essential [32] which is expensive in resource-limited settings like Ethiopia. Hence, measuring the prevalence and directing this scarce resource towards its prevention is the best solution. There is no comprehensive study conducted in Ethiopia to determine triglyceridemia and other plasma lipid profile abnormalities among people living with DM. Therefore, this meta-analysis intended to determine the pooled prevalence of dyslipidemia (high TG, low HDL-C, high LDL-C, and high TC) among people living with DM in Ethiopia.

Methods
2.1. Reporting and Registration. We conducted this systematic review and meta-analysis based on the principles of the Centre for Reviews and Dissemination's (CRD) guidance for undertaking reviews in health care [33], and the reporting system was following the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) [34]. It is registered at Prospero with an identification number CRD42020182291 available at https:// www.crd.york.ac.uk/prospero/#myprospero.

Data Sources.
Major electronic databases (PubMed, Hinari, African Journals Online, and Google Scholar) were used to access relevant studies. Besides, a direct Google search was used to repossess a few articles after looking at the reference lists of included studies to account for the omission of the studies during searching of electronic databases. To account for unpublished data, we used the institutional repository of Addis Ababa University.

Searching Strategy.
All pocket/primary studies reporting proportion or prevalence of hypertriglyceridemia and/or other plasma lipid abnormalities (abnormally high TC, high LDL-C, and low HDL-C), among people living with either T1DM or T2DM in Ethiopia, were the targets for this review. After identifying entry terms using the MeSH (medical subject heading) browser, relevant terms or phrases were combined using Boolean operators "OR" and "AND" to fit the advanced search by considering the CoCoPo (condition, context, population) principle. We  Setting: pocket studies conducted in Ethiopia or any localities of the country.
The outcome of interest: hypertriglyceridemia and other plasma lipid abnormalities (dyslipidemia) among people living with DM.
Study design types: observational studies (cross-sectional or case-control studies) reporting the prevalence or proportion of people with abnormal plasma lipid concentration.

BioMed Research International
Publication status: both published and unpublished studies were included.
Publication language: English. Publication date: published until 9 May 2020.

Exclusion Criteria.
Case-reports, case-series, letters to the journal editors, communications, and studies conducted on specific population were excluded.
2.6. Study Selection. Three independent authors (BD, DG, and YY) identified articles using electronic databases and other sources. All studies were exported from data sources into Endnote X7 for duplicate removal and citation. Two reviewers (BD and YY) selected relevant studies after evaluating title, abstract, and then full-text review for the abstraction of the data by using inclusion criteria. These reviewers harmonized most disagreements after discussion, and the third reviewer (DG) took part to resolve a few divergences between the two reviewers (BD and YY).
2.7. Quality Assessment of Included Studies. Two authors (HD and YA) investigated the quality of included studies using a modified Hoy et al. quality assessment tool [35]. This tool has 9 risks of bias items (each comprises zero or one risk of bias points) which have a maximum score of "9" and a minimum score of "0." Ranking of risk of bias is labelled as low risk (0-3), moderate risk (4-6), and finally high risk (7)(8)(9). The third reviewer (MA) synchronized differences in quality appraisal between the two authors (HD and YA).

Data
Abstraction. Structured data extraction sheet was prepared by two authors (HD and DAA) using Microsoft Excel 2010. The template of data extraction format contained the name of the primary author, publication year of the study, study design, regional state of the study, sample size, actual number of cases with specific plasma lipid abnormality, and percentage of individuals with specific lipid abnormality. Then, two authors (BD and YY) extracted the relevant data independently, and the extracted data of the two authors (BD and YY) were compared. Discrepancies were resolved by consensus after discussion.  participant. Dyslipidemia is defined based on National Cholesterol Education Program Adult Treatment Panel III guideline which comprised either single or a combination of plasma lipid abnormalities. The following criteria were used for the diagnosis of overall dyslipidemia, i.e., TC ≥ 200 mg/ dl, HDL − c < 40 mg/dl for males and < 50 mg/dl for females, LDL − c ≥ 100 mg/dl, and TG ≥ 150 mg/dl [36].
2.10. Statistical Analysis. Endnote X7 and Stata 11 were used for bibliographical management and statistical analysis, respectively. Heterogeneity was tested using the Q statistic, and I 2 test was used to identify possible interstudy variations. The I 2 values 25%, 50%, and 75% denote low, moderate, and high heterogeneity, respectively [37]. We used the DerSimonian and Laird random-effects model [38] for the main metaanalysis of the pooled estimates of the prevalence of hypertriglyceridemia and other plasma lipid abnormalities. Subgroup analysis was considered to account for possible heterogeneity. Publication bias (using the Funnel plot and Egger's test) [39], trim and fill [40], and sensitivity analyses were performed. For each plasma lipid abnormality, the point estimate was shown by the forest plot.

The Pooled Prevalence of Hypertriglyceridemia among
People Living with DM. In the fixed-effects model analysis, the pooled prevalence of hypertriglyceridemia was 43.9% (95% CI: 42.65-45.16%, prevalent cases = 2298) with substantial interstudy variation (heterogeneity). To account for this, we performed the main meta-analysis using the DerSimonian and Laird random-effects model to determine the pooled prevalence of hypertriglyceridemia, and it was 48.15% (95% CI: 38.15-58.15) ( Figure 2).

Other Dyslipidemia among People Living with DM in
Ethiopia. Thirteen studies (3328 participants) were eligible for the estimation of the prevalence of total cholesterolemia among people living with DM. After running the main meta-analysis using the random-effects model, the pooled prevalence of high TC was 34.08% (prevalent cases = 1021) ( Figure 3). Eleven primary studies (2659 participants) were included for the meta-analysis of elevated "LDL-C" with the prevalence of 41.13% (prevalent cases = 889) ( Figure 4). Finally, fifteen studies (3871 participants) were eligible for meta-analysis of low "HDL-C," and the pooled prevalence was 44.36% (prevalent cases = 1432) ( Figure 5).

Subgroup Analysis of the Pooled Prevalence of
Hypertriglyceridemia. As one of the handling mechanisms of heterogeneity, we performed subgroup analysis for the pooled prevalence of hypertriglyceridemia by the type of DM, study design, sampling technique, data source, and publication status. Random-effects analysis showed a 57.80% (9% CI: 50.50-65.10) prevalence of hypertriglyceridemia among T2DM patients whereas 48.64% of the pooled prevalence was observed among published studies (Table 2).
3.5. Metaregression. We applied metaregression to identify the source of heterogeneity. Metaregression was performed using a type of DM used as population interest, sampling procedure, data source, study design, and publication status of the included studies. Of these, only a type of DM was picked as a source of heterogeneity (p ≤ 0:05) ( Table 3).
3.6. Publication Bias. As indicated by the funnel plot ( Figure 6), there was an asymmetrical distribution of studies which is an evidence for publication bias. However, when we performed objective publication bias (egger) test, the bias coefficient of Egger's test was 1.255 (p > 0:05) that indicates absence of publication bias in our study.

Sensitivity Analysis.
The leave-one-out method was used to identify a significant single study effect on the combined estimate. As shown in Table 4, no study exerted a significant effect on the overall pooled prevalence of hypertriglyceridemia among people living with DM.

Discussion
Dyslipidemia is one of the leading causes of cardiovascular diseases [56]. Given its public health significance, scholars conducted many pocket studies in different populations. However, there is no comprehensive study in Ethiopia. The main aim of the current systematic review and metaanalysis was to describe the pooled prevalence of plasma lipid abnormalities (TG, TC, LDL-C, and HDL-C) among people living with DM in Ethiopia. Accordingly, the pooled prevalence was 48.15% for TG, 34.08% for TC, 41.13% for LDL-C, and 44.36% for HDL-C. This indicates a presence of a huge number of individuals with dyslipidemia in Ethiopia, which necessitates the health sector stakeholders to design and implement preventive strategies for plasma lipid abnormalities and hence minimizing dyslipidemia-associated complications of DM and improving their quality of life.
In this meta-analysis, we found a substantial interstudy variation anticipated to affect the pooled prevalence of hypertriglyceridemia and other plasma lipid abnormalities. For the sake of handling the heterogeneity, we used random-effects model analysis for the final combined prevalence of each outcome variable. Even after running the main meta-analysis using the random-effects analysis, there was heterogeneity. Besides, subgroup analysis was executed using type of DM, study design, sampling technique, data source, and publication status.
The prevalence of hypertriglyceridemia among T2DM was 57.8% which is similar with studies in Tanzania (53.8%) [57] and India (56.1%) [58] but lower than that in 4 BioMed Research International           [59]. The pooled prevalence was 48.64% among the published studies which is similar with the overall pooled prevalence of the current study (48.15%). Furthermore, metaregression was performed to sort out the real cause of heterogeneity from which the type of DM was significant source of heterogeneity.
In the current study, the pooled prevalence of hypertriglyceridemia among DM patients was 48.15%. This higher percentage could be related with the defects of insulin action [60,61], in that low insulin or insulin resistance leads to reduced fatty acid mobilization into cells and increases lipolysis, both of which contributed to high plasma lipid levels [62]. Other studies in in Thailand (49.94%) [28], Botswana (38.9%) [63], Tanzania (53.8%) [57], India (56.1%) [58], Brazil (46.7%) [64], and Yemen (39.2%) [65] also reported similar findings. However, the pooled prevalence of   Figure 6: Funnel plot to show publication bias for triglyceridemia. 8 BioMed Research International hypertriglyceridemia, in this study, is lower than a finding of the local study in Jordan (83.1%) [59]. This difference might be due to the variations in the study setting and population characteristics. In contrary to this, our study finding is higher than other reports in the United States (30%) [66], Saudi Arabia (17%) [67], China (22.3%) [68], and Korea (28.7%) [69]. The reason for this might emanate from the differences in sample size, treatment options, adherence to self-care, and duration of DM. In this review, the pooled prevalence of plasma TC was 34.08% which is in line with findings in Botswana (33.5%) [63] and India (36.3%) [58]. However, it is lower than a country-based study in Thailand (35.05%) [28] and local studies in Tanzania (49.6%) [57], Yemen (52.5%) [65], and Jordan (77.2%) [59]. This disagreement might be explained by the differences in the study population in which our study included both type 1 and 2 DM patients, whereas other studies included only people living with T2DM. It is known that people living with T2DM are more prone to plasma lipid abnormalities than T1DM [70]. It can also be due to the differences in the duration of DM, sample size, and treatment adherence. However, the current estimate of TC is higher than nation-wide studies in China [68], another study in China (6.9%) [71], Saudi Arabia (13.8%) [67], and Korea (14.5%) [69]. This might be due to lifestyle and socioeconomic differences between the study populations.
The pooled prevalence of high LDL-C was 41.13%, which is akin to an evidence in the United States (53%) [66]. However, LDL-C in our study is lower than other studies in Thailand (56.54%) [28], Tanzania (72.3%) [57], India (57.3%) [58], Brazil (79%) [64], Yemen (67.5%) [65], and Jordan (91.5%) [59]. The reason for the lower estimated high LDL-C in our review could be due to variations in sample size and scope of the study. However, the estimate of elevated LDL-C is higher than that of other studies in China [68], Saudi Arabia (12.85) [67], and Korea (14.8%) [69]. This could be due to variations in the number of participants, self-management, and duration of DM.
Finally, we estimated the prevalence of abnormally low HDL-C (44.36%) among people living with DM which is in line with other nation-wide studies in Saudi Arabia (40%) [67], Korea (41.2%) [69], and China (40.8%) [68] and local studies in India (35.7%) [58] and Brazil (34.9%) [64]. The estimate of low HDL-C is higher than studies in the United States (23%) [66] and Yemen (25.5%) [65]. This might be due to differences in the knowledge level of participants, self-care practice, and case definition used in each study. Participants in the United States are more likely to have better self-care practice. Our review finding is lower than studies conducted in Thailand (59.59%) [28], Jordan (83.9%) [59], and Tanzania (63%) [57]. This difference could be due to variations in the study population, socioeconomic status, and duration of DM.
As stated, this meta-analysis focused on pooling the prevalence of hypertriglyceridemia and other lipid abnormalities due to lack of studies which reported associated factors of dyslipidemia. In other studies, dyslipidemia is associated with increased age, urbanization, female sex, higher glucose level, smoking, and hypertension [72,73], all of which are modifiable. The findings of this review suggest a need for comprehensive strategies to prevent plasma lipid abnormalities so that dyslipidemia-associated metabolic complications of DM can be avoided by focusing on modifiable risk factors such as lifestyle changes to achieve standard targets. In addition to this, we recommend researchers to conduct a study on determinant factors of dyslipidemia among DM patients so that strategies will be built by targeting modifiable risk factors.

Limitations of the Study.
This review was the first in its kind on this topic and conducted with a robust contribution of different disciplines. However, the finding of this study should be interpreted with the following limitations. Firstly, the total participants included were not a representative of the national figure which might be difficult to generalize. But, it can be used as a good finding than other primary studies since it was a pooled prevalence of others. Secondly, there was high heterogeneity. We performed the random-effects, subgroup analysis, and metaregression to chase for the real cause of the variation, and a type of DM was the cause for heterogeneity. Thirdly, there was lack of studies in certain regions of the country and each year which might be difficult to use the results as a baseline in those regions. Finally, the included studies lack associated factors of dyslipidemia that impeded us to analyze the pooled effect of individual factors on plasma lipid abnormalities.

Conclusions
Our systematic review and meta-analysis revealed a high estimated prevalence of plasma lipid abnormalities, particularly TG among people living with DM. Patients need to adhere to both nonpharmacologic and pharmacologic therapies to reduce abnormal plasma lipid levels and prevent dyslipidemia-associated health risks of DM. Opportunistic test strategies to dyslipidemia parameters have to be strictly followed by all people living with DM at follow-up clinics.

Data Availability
Dataset can be accessed from the corresponding author upon reasonable request.

Conflicts of Interest
All authors declared that there is no competing interest.