Discussion
This is the first study comparing 10-year cardiovascular risk
predictions of Framingham and WHO/ISH risk models for agreement among
Sri Lankans and adds to the limited literature from South Asia and
South-East Asian Region (SEAR) – B. We observed that FRS and WHO/ISH
(SEAR-B) models were not in good agreement in predicting high-risk
patients. Therefore, the FRS and WHO/ISH risk charts should not be used
interchangeably in risk stratification of individual patients during
follow-ups and the same risk model should be used in serial follow-ups
of individual patients of South Asia and South-East Asian Region(SEAR).
Furthermore, we observed that the low information risk models, which do
not need cholesterol values in risk estimation (e.g. FRS with BMI-and
WHO/ISH without-cholesterol) were, good in agreement with standard risk
models using cholesterol values (e.g. FRS with-cholesterol and WHO/ISH
with-cholesterol) in risk stratification of Sri Lankans into low and
high-risk groups. This was also observed in a study of South Asians
living in Canada (except in men aged 60-74 years) where they compared
FRS BMI-based and cholesterol-based models 26.
Therefore, the low information models can be used in the risk
stratification of patients in poor resource settings where laboratory
facilities are space to measurer cholesterol levels among Sri Lankans
and South Asians. However, the dissimilarities reported need be
interpreted cautiously as FRS predicts the risk of all fatal and
non-fatal CVDs including coronary, cerebrovascular, and peripheral
arterial disease and heart failure while WHO/ISH score predicts only
fatal and non-fatal myocardial infarctions and strokes and therefore the
two tools are not directly comparable.
Ranawaka et al. 27 studied cardiovascular risk
estimates of a Sri Lankan community, in 2007 and observed 8.2%
prevalence of high-risk patients using the WHO/ISH model in a cohort of
patients with and without previous CVDs. We studied a cohort of Sri
Lankans without previous CVDs in 2019 and observed 9.5% of them being
at high-risk the with WHO/ISH(-with cholesterol) model. Therefore, our
findings seem consistent with previous literature. Ranawaka et al.
compared risk predictions of WHO/ISH, National Cholesterol Education
Program - Adult Treatment Panel III (NCEP-ATP III) and Systematic
Coronary Risk Evaluation (SCORE) models and observed a difference
between the predictions of the three models; 8.2%, 25.4%, and 11.8%
respectively being categorised as high risk. NCEP-ATP III model, which
is a derivative of FRS, predicted 25.4% of the Sri Lankan being at high
risk, compared to 8.4% predictions with WHO/ISH model, which was much
higher than the prediction of NCEP-ATP III model. We also observed that
the predictions of FRS were higher than that of WHO/ISH models.
Results of other Asian studies comparing of Framingham score and WHO/ISH
charts also reports the two being different but the literature on best
risk estimates for South Asians is not consistent. Several studies
identified WHO/ISH score underestimating the risk of Asians compared to
FRS 28-30. It was reported that The FRS and SCORE-high
models, but not the WHO/ISH model can be used to identify high
cardiovascular risk in Malaysians 28. The same was
reported by a few other Asian studies of Cambodia, Mongolia, Malaysia,
and Jamaica 31,32. Few Indian studies reported the FRS
CV risk assessment model has performed the best to identify patients at
high CVD risk while WHO and ASCVD calculators were the worst 29,30,33. In contrast, Asia Pacific Cohort Studies
Collaboration observed that FRS overestimated CV-risk in Chinese but was
adequately predictive when it was recalibrated using contemporary data.
However, this study did not assess CV-risk with WHO/ISH r charts12. Further to that, some reported that the Framingham
and British scores underestimate CVD risk in Asian Indians and the need
for developing specific models for them 34.
CV-risk assessment should be based ideally on data of epidemiological
risk factors appropriate to the population to which it is applied to35. WHO/ISH charts were developed using extrapolated
data on CV risk factors in different geographical regions but have not
been systematically validated prospectively in most populations and
therefore may perform poorly 5. Framingham score may
not be reflecting CV-risk of Asians accurately due to several reasons.
Framingham risk score was developed using data of Americans in an era
when the CVD risk was very high in the USA but was less in Asia18. It does not take into account some important risk
factors relevant to Asians like abdominal obesity, physical inactivity,
and family history of premature CVD, which are currently increasing in
prevalence among Asians. The differential effect of ethnic groups,
environments, and genes on the risk of CVD could also play a part36,37. Also, Asians are relatively younger when they
develop CVDs compared to white Caucasians 12,13 while
“age” is a very strong risk factor in the Framingham model. Therefore,
FRS may underestimate mate CV-risk of young Asian patients. Therefore,
the best model to risk stratify South Asians is still a query.
After all, a screening tool should be able to detect all patients at
high-risk without missing high-risk patients while having an acceptable
false positive detection rate. The prevalence of high-risk patients
according to WHO/ISH (SEAR – B) model is much lower than that with the
FRS model and therefore, maybe a lot of high-risk patients fromSEAR – B
region are not getting adequate primary prevention measures, due to
being risk-stratified with WHO/ISH (SEAR – B) model. We cannot be
certain until a new model is developed of a cohort of South Asians or
the existing risk scores are validated in them.
Our study has many strengths. This is a very thorough study with
complete data. We used only the patients who did not have any previous
CVDs. We have calculated risk predictions of a given patient using 4
models and compared them in pairs and therefore there is no selection
bias. However, there are a few limitations of our study as well. This is
not a random sample of Sri Lanka and is of a high-risk population and
therefore, the rates of risk factors prevalence and mean risk scores of
this cohort may not be generalizable Sri Lanka. However, the conclusion
that “FRS and WHO/ISH charts are only in satisfactory agreement in the
prediction of high-risks Sri Lankans ” is generalizable, as we arrived
at this conclusion by comparing risk scores within each patient,
nullifying any selection bias.