CARDIOVASCULAR JOURNAL OF AFRICA • Vol 24, No 8, September 2013
AFRICA
331
targets.
21-23
However, studies from all over the world have
demonstrated that many patients on lipid-lowering therapy do
not reach their recommended lipid targets.
24-26
The South African
Heart Association (SA Heart) together with the Lipid and
Atherosclerosis Society of Southern Africa (LASSA) therefore
recently emphasised that intensive management of dyslipidaemia
could significantly reduce the South African CVD health
burden.
21
The DYSlipidaemia International Study (DYSIS) is a cross-
sectional, observational study that has examined the efficacy of
lipid-lowering therapies in patients from various regions of the
world, including Canada and Europe (11 countries), in order
to better characterise predictive factors for dyslipidaemia and
CVD.
24,25
Here, as part of DYSIS, we have analysed residual
dyslipidaemia in statin-treated South African patients.
Methods
As part of DYSIS, this epidemiological, observational, cross-
sectional study was conducted in South Africa between 1
November and 9 December 2011. Data for the study were
collected in the South African private healthcare sector by 16
physicians; 50% were primary-care physicians and 50% were
specialised office-based physicians (e.g. cardiologists).
Prior to study initiation, the relevant local ethical review
committees approved the study protocol and all patients gave
written informed consent before enrolling in the study. Key
eligibility criteria were: (1) age of at least 45 years, (2) receiving
stable statin therapy for at least 3 months, and (3) fasting for
at least 12 hours at the time of visit while on statin therapy.
Participating physicians were instructed to include all eligible
and consenting patients consecutively.
Patient demographic, lifestyle and clinical characteristics were
documented. Lipid levels (total cholesterol, LDL-C, HDL-C and
triglycerides) were measured using the CardioChek
®
device
) at the time of patient enrollment
to reliably collect lipid measurements uniformly at all sites. The
LDL-C test strip provided measures LDL-C directly across a
range of 1.29–5.18 mmol/l in about two minutes.
Additionally, the lipid-lowering regimen at the time of the
most recent blood sample was recorded for each patient (in
particular, statin type and daily dose) as well as any information
regarding other lipid-modifying therapies. The potency of
different types of statins was normalised using a calculation that
allows benchmarking against six different simvastatin dose levels
(5, 10, 20, 40, 80 and 160 mg/day), with potency scores ranging
from 1 (5 mg/day simvastatin) to 6 (160 mg/day simvastatin).
23,27
The 2011 ESC guidelines were used to classify CV risk,
LDL-C level treatment goals, and sub-optimal HDL-C and
triglyceride levels.
21,28
Variables independently associated with
dyslipidaemia were evaluated with logistic regression modelling
using the following variables: age (
≥
70 years), female gender,
family history of premature coronary heart disease (CHD),
current tobacco smoker, sedentary lifestyle, alcohol consumption
(
>
2 units/week), body mass index (BMI)
≥
30 kg/m
2
, large
waist circumference (
>
102 cm in men,
>
88 cm in women
29
),
hypertension, DM, coronary heart disease, cerebrovascular
disease, heart failure, peripheral artery disease, systolic/diastolic
blood pressure
≥
140/90 mmHg, simvastatin equivalent dose
of either 20 to 40 versus 10 mg/day, or
>
40 mg versus 10 mg/
day, ezetimibe use, and physician’s specialty (cardiologist,
endocrinologist, diabetologist, internal medicine or other).
Statistical analysis
To estimate the sample size needed for South Africa we assumed
a prevalence of residual lipid abnormalities between 20 and
60% in patients fulfilling the entry criteria for this study and a
design effect of 20% (variance inflation due to cluster sampling
design). We calculated that, within this range, a sample size of
1 000 would be sufficient to estimate the prevalence of residual
dyslipidaemia with a given precision of
±
3.4% (range of 95%
confidence interval: 6.8%). Furthermore we determined that
this size guaranteed enough information for estimating the
prevalence in smaller subgroups (representing one-quarter or
more of the population) with a precision of
±
6.8% (95% CI:
13.6%).
Following data collection, patient information was entered
into a central web-based database housed and managed at
the Institut für Herzinfarktforschung, Ludwigshafen, Germany.
Real-time quality control (internal logic checks) occurred during
web-based data entry. Continuous variables are presented as
means with standard deviations or medians with 25th and
75th percentiles [interquartile range (IQR)] as indicated, and
categorical variables are reported as absolute numbers and
percentages.
Kernel density estimation was used to analyse the distribution
of total cholesterol, LDL-C, HDL-C and triglyceride levels.
The value of a kernel density and its slope at the lipid value
equal to the ESC goal provides a crude indicator of the change
in the proportions of patients meeting the goal from a small
improvement or deterioration in lipid level starting from the ESC
goal. This approach thus provides a sensitivity analysis for either
changes in the ESC goals or changes in lipid levels for people
whose levels are near the goals.
Multiple logistic regression analyses with backward selection
(
α
=
0.05) were used to identify variables independently
associated with LDL-C, HDL-C and triglyceride irregularities.
Two-tailed statistical comparisons were used (
p
<
0.05 was
significant) and patients lacking the appropriate lipid parameters
were not included within the analyses. All analyses were
performed using SAS v 9.1 (SAS Institute Inc, USA).
Results
Patient characteristics, risk categories and lipid parameters are
presented in Table 1. The study enrolled 1 029 patients (429
men, 600 women). The mean age of patients was 65.4 years, and
58.3% were female. The study population was of mixed ethnic
(multi-racial) origin, including Caucasians (56.6%), blacks
(22.0%), Asians (9.5%) and patients of mixed ancestry (12.0%).
Patient characteristics and cardiovascular risk profile differed
by ethnic group. A family history of premature CVD was
reported by 34% of Caucasian patients while the diabetes
prevalence of 25.6% was the lowest of all the ethnic groups
studied. Hypertension was found in 69.8% and CVD in 41.1%
of Caucasian patients. Black patients were least likely (1.8%)
to report a family history of premature CHD and had the
lowest (5.3%) smoking rates. However, hypertension was almost
universal (93.3%) and diabetes and obesity were highly prevalent
at 71.2 and 61.9%, respectively. Despite the high prevalence
