CARDIOVASCULAR JOURNAL OF AFRICA • Vol 23, No 2, March 2012
104
AFRICA
Poor record keeping precludes the use of administrative data-
bases to inform public healthcare policies. Cross-sectional data
relating to the distribution of risk factors and the prevalence of
CVD exist in some places, as summarised elsewhere.
1,7-12
That
the availability of this type of data has not produced the expected
change in policies to counter the trend of CVD, highlights the
need for more sensitive evidence on the ill effects of CVD in
SSA. In the West for example, the observed decline in incidence
of CVD has been largely influenced by evidence generated
from longitudinal studies (interventional or not), including the
landmark Framingham Heart Study initiated around the peak in
incidence of CVD in that part of the world.
Longitudinal studies of cardiovascular diseases in Africa
have several applications, including: (1) generating more sensi-
tive information in the form of causal associations between risk
factors and hard outcomes such as death and disability, and
therefore increasing awareness and need for action; (2) contex-
tualising the knowledge generated elsewhere on CVD, and
accordingly, improving the local uptake of measures with proven
benefits on cardiovascular outcomes in other parts of the world;
(3) providing the unique opportunity of accurately characterising
the early phases of epidemiological transition, and the interaction
between CVD and prevalent infectious diseases; (4) providing
local epidemiological training laboratories to mould the careers
of many young African researchers to continue the fight against
CVD across the generations; and (5) providing resources for
collaboration between African researchers and their peers with
similar interests around the world.
The study aim was to conduct a systematic review of the
literature for all prospective cohort studies of cardiovascular
traits that have measured exposure before outcome in SSA.
We were interested in identifying gaps in the knowledge on
the epidemiology of CVD in SSA. The objective was to assess
the suitability of the available studies for reliably addressing
research uncertainties through data pooling. Such information
is useful for informing the immediate health services and policy
solutions, and assisting the design and planning of relevant stud-
ies that will inform future strategies.
Methods
Data source
We systematically searched the PubMed and Africa EBSCO
databases, using a strategy that included all possible combina-
tions of three levels of medical subject heading terms: (1) ‘Africa
south of the Sahara’ (2) ‘cohort studies’, ‘longitudinal studies’,
‘retrospective studies’, ‘prospective studies’, and (3) ‘cardio-
vascular diseases’, ‘stroke’, ‘hypertension’, ‘diabetes mellitus’,
‘smoking’ and ‘cholesterol’. The search was limited to studies
in humans. The starting date from which articles were identified
was from 1966 until October 2009.
We searched the database of cohort studies of the
International
Journal of Epidemiology
. References quoted in original publica-
tions, two editions of a book on causes of deaths and diseases in
Africa,
13,14
and the INDEPTH network website
15
were searched
for additional information. The Global Cardiovascular InfoBase
of the World Health Organisation
16
was also consulted. We
limited the review to articles that provided at least an abstract
in the English language. Titles of the articles and abstracts were
reviewed and relevant articles obtained if required. When the full
article was needed and was not available to us, attempts were
made to get one from the corresponding authors. References
were extracted and stored with the use of Endnote V9.0.0 soft-
ware (Thomson/ISI ResearchSoft, Berkeley, CA).
Data selection
Two reviewers (APK and LNM) independently screened the arti-
cles for eligibility. The inclusion criteria were: prospective cohort
design; measurement of exposure before outcome; minimum
duration of follow up of six months; baseline assessment for at
least one major risk factor other than gender and age (i.e. blood
pressure variables/status for hypertension, lipid variables/status
for dyslipidaemia, glucose exposure/status for diabetes, smoking
status) or for a status for cardiovascular disease; and/or outcomes
ascertainment during follow up, including trajectories of risk
factors and mortality; and studies conducted in a sub-Saharan
African country. We excluded migrant studies, studies with a
focus on non-cardiovascular diseases, post-surgical intervention
cohorts, and post-cardiac instrumentation cohorts. Cohorts with
less than 30 participants at baseline were also excluded.
Data extraction
We extracted data on the country of the study, the year of incep-
tion, main focus of the study, number of participants at baseline
and final evaluation, the setting of the study (hospital, commu-
nity, both, other), baseline measurement and outcomes, and
the overall duration of follow up. We did not perform a quality
assessment.
Statistical analysis
To assess the homogeneity between studies, we computed the
ratio of number of participants successfully traced at the final
visit/number of participants assessed at baseline (with the
accompanying 95% confidence intervals) for each study that
provided enough data to compute this ratio. We then constructed
a forest plot of these ratios and the pooled estimate, assum-
ing a random effect model. These analyses were performed
using the Comprehensive Meta Analysis V 2.2.046 (Biostat,
Inc. Englewood, USA) and Meta-analysis with Interactive
Explanations (MIX)
17,18
V 1.7.
Results
The initial search of electronic databases revealed 788 entries
published between 1966 and 2009. Of these references, 676 were
excluded because they were not relevant to the purpose of this
systematic review. A total of 81 references reporting on 41 stud-
ies were included in the final review (Fig. 1). These studies had
been conducted in 11 sub-Saharan African countries, with about
59% of them in South Africa and Nigeria.
Articles relating to the same study were grouped and checked
for consistency. Studies were hospital-based historical or
prospective cohorts (22 studies) or community-based cohorts
(10 studies). Few had a hospital and community component and
work place-based cohorts (two studies). Few studies were still
ongoing and others were conducted over a range of duration from
six months to over 20 years.
The focus of these studies varied substantially, with a concen-
