CARDIOVASCULAR JOURNAL OF AFRICA • Volume 32, No 3, May/June 2021
138
AFRICA
(9.6%) (Table 1) falls within the range of values reported in a
systematic review and meta-analysis comprising 23 studies and
over 15 000 participants, where the respective FMD values varied
between 2.3 and 13.8%.
12
In our study, females had significantly
higher FMD% values compared to males (Table 1). This agrees
with findings from the Framingham Heart Study, which showed
similar trends in their cohort,
34
and emphasises the importance
of taking gender differences into account when measuring
FMD in study populations. However, this phenomenon may be
explained by the significantly larger baseline artery diameters
measured in the male participants compared to females, which
has also been shown by others.
16,34
Body weight parameters appeared to be associated with
FMD%, as suggested by a modest correlation with waist
circumference (Table 2), and observing higher FMD% values
in participants with BMI
≥
25 kg/m
2
. However, there was no
correlation when waist circumference was expressed as a ratio
of hip circumference (WHR) (Table 2) and the age-adjusted
association with overweight/obesity was lost when additionally
adjusting for gender. Furthermore, the presence of central
obesity had no effect on FMD% and no correlation was
observed between BMI and FMD%. Taken together, the results
show that body weight parameters were not strongly associated
with FMD%, likely due to the relatively young mean age of the
cohort and the fact that waist circumference, WHR and BMI
values were well within the normal ranges.
A previous study by Brook
et al.
35
also failed to show a
significant relationship between BMI and FMD%, while noting
that the WHR was inversely associated with FMD%. Although
another study showed an inverse association between BMI and
FMD%, the study population included subjects with only severe
obesity,
34
which was not the case in our cohort. Participants with
high total cholesterol levels (
>
5.1 mmol/l) had increased FMD%
values, however, this age-adjusted relationship disappeared when
the model was additionally adjusted for gender, and was therefore
unlikely to be of physiological relevance. None of the other lipid
and cardiometabolic variables showed relationships with FMD.
In the present study cohort, mean CRAE values corresponded
reasonably well with those reported in a systematic review
and meta-analysis by Ding
et al.
36
comprising over 10 000
participants, and studies conducted in the North West Province
of South Africa.
37,38
The median CRVE value in our cohort
(253.6
µ
m) (Table 1) was higher than the range of CRVE values
(192.3–231.2
µ
m) reported by Ding
et al.
36
On the other hand,
the present cohort’s CRVE compared reasonably well with that
of a study in a cohort from the North West Province in South
Africa,
38
and tended to be lower than that measured in one of
the only other Western Cape Province-based studies investigating
retinal microvascular calibres in a Cape Town population.
21
The apparent inter-study inconsistency may be related to
differences in retinal image analysis procedures and methodology,
including different software packages used for analysis, as
reported previously.
39
In addition, the discrepancy may also be
ascribed to our cohort’s relatively young mean age of around 35
years, compared to a mean age range of between 50 and 61 years
in the meta-analysis study of Ding
et al
.,
36
as previous reports
have shown that retinal venular diameters narrow with increasing
age.
40
The higher smoking prevalence in our cohort (~86%)
(Table 1) compared to the smoking rates in the studies reviewed
by Ding
et al
.
36
(11.5–23.7%) may also explain the higher CRVE,
since cigarette smoking has been shown to be strongly associated
with wider retinal venular calibres.
20
In our cohort, there was an inverse correlation between
CRAE and systolic and diastolic blood pressure (Table 2). In
addition, participants with systolic hypertension had lower
CRAE values compared to normotensive participants (Fig.
2A), which is in accordance with findings by others, suggesting
that narrower retinal arteriolar width may be a marker of
hypertension.
20,36,41,42
This finding underscores the potential value
of retinal screening in young adults, such as the current
cohort, who may not yet clinically present with hypertension.
Results also show decreased CRVE values in participants with
systolic and diastolic hypertension compared to normotensive
participants (Fig. 2B), which is in agreement with observations
made in a previous study.
43
However, the relationship between hypertension and retinal
venular narrowing is controversial, with some authors arguing
that venular narrowing may be confounded by concomitant
arteriolar narrowing.
44
In the present cohort, additional
adjustment for CRAE resulted in a loss of significance in the
systolic hypertension model, while a borderline significance
was maintained in the diastolic hypertension model. Fasting
triglyceride levels also showed a significant relationship with
wider retinal venules (Table 2), as previously shown by others.
42
To further investigate the presence of abnormal retinal
features and their potential association with cardiometabolic
variables, the retinal images of the present cohort were subjected
to qualitative fundus grading, which identified retinal tortuosity
(mostly arteriolar) in around 18% of the participants. Results
showed that participants with retinal tortuosity had increased
diastolic blood pressure compared to participants with no signs
of tortuosity. Retinal arteriolar tortuosity has previously been
associated with increased systolic and diastolic blood pressure.
45,46
The demonstration of an association between retinal
microvascular changes and endothelial dysfunction of systemic
arteries has important implications, as it may provide an
opportunity to use the retinal microvasculature as a surrogate
marker of systemic vascular disease. We could not demonstrate an
association between CRAE or CRVE and FMD% in our cohort.
In the literature, the relationship between retinal microvascular
calibre and FMD% remains generally inconclusive, however,
one previous study did demonstrate an independent association
between CRVE and systemic endothelial dysfunction as measured
by brachial FMD.
33
Limitations
The study has shortcomings that need to be considered when
interpreting the findings. This was intended to be a pilot study
to obtain baseline data in a generally disease-free group of
participants, hence the relatively small sample size. This placed
limitations on some of the statistical analyses, where a number
of borderline significant
p
-values (0.05–0.08) were noted. A
larger sample size may have generated more significant outcomes
and allowed for the inclusion of more sophisticated and robust
association analyses such as multiple regression models.
Furthermore, it has to be acknowledged that the cohort was
recruited from a relatively restricted geographical location with
limited demographic variability, which limits the extent to which
the data can be regarded as representative of the wider South