Unmet need for hypercholesterolemia care in 35 low- and middle-income countries: A cross-sectional study of nationally representative surveys

Background As the prevalence of hypercholesterolemia is increasing in low- and middle-income countries (LMICs), detailed evidence is urgently needed to guide the response of health systems to this epidemic. This study sought to quantify unmet need for hypercholesterolemia care among adults in 35 LMICs. Methods and findings We pooled individual-level data from 129,040 respondents aged 15 years and older from 35 nationally representative surveys conducted between 2009 and 2018. Hypercholesterolemia care was quantified using cascade of care analyses in the pooled sample and by region, country income group, and country. Hypercholesterolemia was defined as (i) total cholesterol (TC) ≥240 mg/dL or self-reported lipid-lowering medication use and, alternatively, as (ii) low-density lipoprotein cholesterol (LDL-C) ≥160 mg/dL or self-reported lipid-lowering medication use. Stages of the care cascade for hypercholesterolemia were defined as follows: screened (prior to the survey), aware of diagnosis, treated (lifestyle advice and/or medication), and controlled (TC <200 mg/dL or LDL-C <130 mg/dL). We further estimated how age, sex, education, body mass index (BMI), current smoking, having diabetes, and having hypertension are associated with cascade progression using modified Poisson regression models with survey fixed effects. High TC prevalence was 7.1% (95% CI: 6.8% to 7.4%), and high LDL-C prevalence was 7.5% (95% CI: 7.1% to 7.9%). The cascade analysis showed that 43% (95% CI: 40% to 45%) of study participants with high TC and 47% (95% CI: 44% to 50%) with high LDL-C ever had their cholesterol measured prior to the survey. About 31% (95% CI: 29% to 33%) and 36% (95% CI: 33% to 38%) were aware of their diagnosis; 29% (95% CI: 28% to 31%) and 33% (95% CI: 31% to 36%) were treated; 7% (95% CI: 6% to 9%) and 19% (95% CI: 18% to 21%) were controlled. We found substantial heterogeneity in cascade performance across countries and higher performances in upper-middle-income countries and the Eastern Mediterranean, Europe, and Americas. Lipid screening was significantly associated with older age, female sex, higher education, higher BMI, comorbid diagnosis of diabetes, and comorbid diagnosis of hypertension. Awareness of diagnosis was significantly associated with older age, higher BMI, comorbid diagnosis of diabetes, and comorbid diagnosis of hypertension. Lastly, treatment of hypercholesterolemia was significantly associated with comorbid hypertension and diabetes, and control of lipid measures with comorbid diabetes. The main limitations of this study are a potential recall bias in self-reported information on received health services as well as diminished comparability due to varying survey years and varying lipid guideline application across country and clinical settings. Conclusions Cascade performance was poor across all stages, indicating large unmet need for hypercholesterolemia care in this sample of LMICs—calling for greater policy and research attention toward this cardiovascular disease (CVD) risk factor and highlighting opportunities for improved prevention of CVD.


Introduction
• gap is overstated: authors make two seemingly contradictory points in the introduction: 1) dyslipidemia is widely prevalent in LMIC (cite literature including GBD reports), and 2) "literature has been unable to provide accurate country prevalence estimates due to dearth of nationally representative data" o GBD uses WHO Steps data which comprises part of the authors dataset o authors overstate their claim that "literature has been unable to provide accurate country prevalence estimates" ▪ in fact, multiple GBD publications that the authors cite do exactly this, generate reliable population level estimates • last sentence first paragraph-statement that high burden of dyslipidemia persist b/c of emphasis on communicable, maternal, and neonatal diseases…. That seems to be speculation. Could be due to lack of MOH funding for lab capacity that includes lipids (usually a send out lab in most LIC) for example. Sets up a narrative that may not have empirical evidence.

Methods
• age for analysis: starting at 15 too young o in most countries, unless you had a genetic condition like familial hyperlipidemia, people < 40 yrs with elevated LDLc would not qualify for a statin o restricting it to people aged 40 years seems to be more appropriate, especially if authors do not want to avoid evaluating health systems by protocols that are not standard of care ▪ they've partly done this by reporting cascade for individuals meeting PEN criteria for lipid screening • Need clearer definitions of each step-do they include a time period-for example, lipids measured within the last year or within the last 10 years? • What is the definition of HTN-SBP 130 or 140 for all comers, for those with DM/CRF? Definition of DM also not provided.

Discussion
• The main points per the results appear to be overall prevalence of dyslipidemia was ~7% and the major drop using cumulative proportions is lipids measured (43-50%)….the other steps are actually much higher. Yet the discussions starts with the points of treatment and control which seems to be secondary points. • Moreover, the gap in paragraph 2 seems to be ~ to high income countries which is not given enough discussion as to why. • Overall the discussion is very long and very unfocused. For example, authors dedicate 1-2 paragraphs to each of screening, diagnosis, treatment, and control. These paragraphs have two components: repeating data from this study (and stating there is heterogeneity), and comparing this to data from other care cascade studies. Because of this structure, there is a lot of repetitiveness in 1) repeating data from study, and 2) repeated comparisons with other countries. Examples below: • repeating data from this study: o While there was substantial heterogeneity across countrieswith screening rates ranging from under 10% in Benin, Kiribati, Myanmar, the Solomon Islands, and Zambia up to 89% in Iranthis stage was consistently found to be the largest or second largest stage of loss along the cascade of care. o Again, large heterogeneities across the sample became apparent at this stage, as less than 10% of individuals with dyslipidemia in six countries reached this stage, less than 50% in 23 additional countries, and only six countries were able to retain more than 50% of individuals with dyslipidemia at the diagnosis stage. o We found that 29% of those with high TC and 33% of those with high LDL-C had been given advice or medication for their high cholesterol. At the country level, this ranged from less than 5% up to 79%. o Both in the pooled analysis and at the disaggregated level, control rates were found to be low, ranging from virtually zero to 27% in all but two countries -Iran (57%) and Morocco (49% • this sentence is confusing, pages 7-8: "Second, as treatment guidelines are usually based on CVD risk scores, rather than on lipid measures alone, and vary across countries, we chose to be conservative in our definition of who is considered to be in need of care as to not evaluate health systems by unapplied care standards"

Methods
• how was missing data handled? requirement for surveys to have at least 50% survey response to be included in pooled data sets, but that still could be a lot of people missing o can authors clarify which weights they used from each country's survey weights? Some have multiple depending on which data is examined (interview vs biochemical). This would help with transparency and reproducibility o how did authors adjust for missing data? did each country's survey weights adjust for missing data? • page 7: can authors define cutoff values for the 3 classifications of total cholesterol and 5 classifications for LDLc • page 11: last paragraph in Statistical Analysis section doesn't seem to belong there. Seems better suited to a Measurements section.

Results
• given 41% sample is < 35 years, should probably report age standardized dyslipidemia prevalence too • Fig 1: font very small, hard to read • page 12 type: Easter Mediterranean • "multivariate Poisson regression" should be "multivariable Poisson regression". See https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3518362/. • page 14, "less than two out of every five respondents with dyslipidemia had been treated…" but in Fig 1 the