• Open Access

Markedly different clustering of CVD risk factors in New Zealand Indian and European people but similar risk scores (PREDICT-14)


Correspondence to: Dr Lavinia Perumal, Auckland Regional Public Health Service, Private Bag 92605, Symonds Street, Auckland, New Zealand; e-mail: laviniap@adhb.govt.nz


Objective: To compare the cardiovascular disease (CVD) risk profiles of Indian and European patients from routine primary care assessments in the northern region of New Zealand.

Method: Anonymous CVD risk profiles were extracted from PREDICT (a web-based decision support program) for Indian and European patients aged 35–74 years. Linear regression models were used to obtain mean differences adjusted for age, gender and deprivation.

Results: At recruitment, Indian participants (n=8,830) were younger than Europeans (n=47,091), in keeping with national guidelines that recommend earlier CVD risk assessment for Indians. Compared with Europeans, a greater proportion of Indian participants lived in areas of higher deprivation and had a two to four-fold greater burden of diabetes in all age groups. Indian participants had a significantly lower proportion of smokers and a lower mean systolic blood pressure. The respective cardiovascular risk factor profiles lead to similar age-adjusted Framingham five-year CVD risk scores.

Conclusions and implications: National data sources indicate that there are higher rates of hospitalisations and deaths from CVD in Indians compared with Europeans. Our study found similar predicted CVD risk in these two populations despite markedly different clustering of risk factors, suggesting that the Framingham risk equation may underestimate risk in Indians. There is a need for better ethnicity coding to identify all South Asian ethnicities.

Despite the recent decline in hospitalisations and mortality from cardiovascular disease (CVD) in the New Zealand population, certain ethnic groups continue to have a disproportionately high burden. The New Zealand CVD risk assessment guideline1 specifies people of South Asian ethnicity as a high-risk population group. Here, South Asians include the following ethnicities: Indian, Sri Lankan, Afghani, Bangladeshi, Nepalese, Pakistani and Tibetan.

Knowledge of CVD risk distributions and outcomes for South Asians in New Zealand is limited and mainly summarised in two publications, the Asian Health Chart Book2 and Asian Health in Aotearoa.3 The former collated data relating to Indian people and the latter for Indian and Sri Lankan. Compared with European and other Asian groups, the reports indicated higher hospitalisation and mortality rates2 for ischaemic heart disease (IHD) and higher self-reported IHD and diabetes3 among South Asian communities.

PREDICT is a web-based computerised decision support system that assists with primary care CVD risk assessment during patient consultation. The tool is integrated with commonly used patient management systems. This integration allows systematically coded CVD risk data to be automatically extracted from a patient's electronic medical record. Any gaps in the required CVD risk data can be filled in by the GP or practice nurse. Risk profiles are sent via secure broadband internet connection to a central server. Within seconds the clinician receives the patient's calculated five-year CVD risk as well as risk management recommendations. Primary care practices conducted CVD risk assessments mainly opportunistically and gave permission for anonymised patient data to be aggregated for research. PREDICT is currently used by more than 1,500 primary care practitioners. We investigated differences in CVD risk profiles between Indian and European participants at their first assessment.


We analysed the results of PREDICT assessments from August 2002 to November 2009 from consenting practices in the Auckland and Northland regions. The PREDICT templates allow standard CVD risk assessment data (age, gender, ethnicity, prior history of CVD, diabetes, family history of premature CVD, smoking, blood pressure and lipid profile) to be recorded. Data definitions have been described previously.4 The Framingham CVD equation was used to estimate five-year CVD risk after excluding those with a history of prior CVD or CVD equivalent (diabetes complicated by nephropathy or familial genetic lipid disorder).

Only those aged 35 to 74 years were included in these analyses as: (a) the lower age limit is the recommendation for commencing CVD risk assessment in males from high-risk ethnicities,1 and (b) the Framingham CVD equation has an upper level of 74 years.

PREDICT allows for up to three ethnicity fields to be completed, which were mainly automatically populated from the patient's records. This could be changed at the time of consultation. Demographic data were anonymously linked to the equivalent data held by the New Zealand Health Information Service (NZHIS) in the National Health Index (NHI) database. The NZHIS-NHI dataset also has three ethnicity fields per person. Coding of ethnicity was according to Ministry of Health protocols.5 This study represents a sub-group of the PREDICT database who were able to be identified clearly as to ethnicity, i.e. Maori, Pacific, Indian, European, Other Asian, and Other ethnicities. There is no single ethnicity code that aggregates information for all ‘South Asian’ ethnicities as it consists of a mixture of level 2, 3 and 4 ethnicity codes. Unfortunately, only level 2 ethnicity coding is available from current patient management system software and NZHIS-NHI datasets. This allowed for identification of Indian patients but not other South Asian ethnicities. The European group in this study included participants who were ‘New Zealand European’, ‘Other European’, ‘European not further defined’ or ‘ethnicity not stated’.

New Zealand Deprivation (NZDep) Index scores were ascertained from the NZHIS-NHI dataset. The NZDep is a small area index of socio-economic deprivation that provides a score for each census mesh block based on nine variables from the 2001 Census.

Statistical analysis was performed using Stata version 10.0. Data were stratified by age and gender. Chi-square (χ2) tests were conducted to assess univariate associations. For continuous measures, two sample t-tests were used to calculate mean differences and 95% confidence intervals. Linear regression models were used to obtain mean differences adjusted for age, gender and deprivation.

Ethical approval was obtained in 2003 (AKY/03/12/314) with subsequent approval by the national Multi–Region Ethics Committee in 2007 (MEC/07/19/EXP).


A total of 47,091 Europeans and 8,830 Indian people aged 35 to 74 years had completed PREDICT risk assessments. Indian males were, on average, seven years younger than European males. Indian females were five years younger than European females (Table 1). The Indian population included a greater proportion of males compared with Europeans (61% vs 53%). Almost 60% of Indian people lived in NZDep quintile 4 or 5 (most deprived) areas compared with 42% of Europeans.

Table 1.  Number and proportion of participants by ethnicity, age group and gender.
Age group (years)Indian (n=8,830)  European (n=47,091)  
 Males (n=5,368)Females (n=3,462)Males (n=24,948)Females (n=22,143)
Mean age (SD)49.0 (9.7) 53.9 (8.3) 56.5 (9.6) 59.0 (8.8) 
Median age48 53 5760 

Table 2 presents the distribution of CVD risk factors. There were significantly greater proportions of Indian males and females who had diabetes than Europeans. In each 10-year age and gender group, Indian people had a two to four-fold greater burden of diabetes. Conversely, Indian males and females had significantly lower proportions of people who smoked or who had a family history of premature CVD.

Table 2.  Distributions of cardiovascular risk factors among participants, by ethnicity, age group and gender.
CVD Risk factorAge group (years)Indian Males (n=5,368) European Males (n=24,948) χ2p-valueIndian Females (n=3,462) European Females (n=22,143) χ2p-value
  n%n% n%n% 
Prior history of CVD35–44361.8853.1<0.00162.2291.9<0.001
 45–541126.15086.6 352.11733.8 
 55–6418117.21,24614.7 656.06647.0 
 65–7412527.51,76629.6 7216.41,05816.4 
 45–5455129.86568.5 39023.547910.4 
 55–6444141.91,22614.5 43439.998110.3 
 65–7423952.51,35022.7 22551.41,08316.8 
 45–5421611.71,41418.3 231.480917.5 
 55–64979.21,24714.7 90.81,21312.8 
 65–74204.463210.6 51.15117.9 
Family history of premature CVD35–4424212.183730.0<0.0019133.050332.6<0.001
 45–5426514.31,84323.9 19811.91,50732.6 
 55–6411811.21,70720.1 12111.12,41525.4 
 65–74306.688514.9 4811.01,37021.2 

When adjusted for age, Indian males and females had significantly lower mean systolic and diastolic blood pressures (Table 3). In contrast, no clinically significant differences were seen in mean total cholesterol to high density lipoprotein ratio (TC/HDL).

Table 3.  Mean differences for blood pressures and TC/HDL ratio by ethnicity, stratified by age group and gender. Thumbnail image of

The mean Framingham five-year CVD risk scores for Indian and European participants (without prior history of CVD or CVD equivalent condition) were similar. The overall age-adjusted mean CVD five-year risk scores for Indian males was 0.05% higher and Indian females 0.3% lower than Europeans (Table 4). When adjusted for both age and deprivation, mean differences in Framingham risk, blood pressure and TC/HDL were not significantly different.

Table 4.  Mean Framingham five-year CVD risk scores for participants by ethnicity, age group and gender.
GenderAge group (years)Indian Mean (SD)European Mean (SD)Mean difference (95% CI)T-test (p-value)
Male35–441.8 (1.7)2.6 (2.1)–0.7 (–0.8, –0.6)<0.001
 45–544.8 (3.4)5.2 (3.6)–0.4 (–0.6, –0.2)<0.001
 55–649.0 (4.9)9.4 (5.2)–0.3 (–0.7, 0.04)0.079
 65–7413.8 (6.4)13.8 (6.2)–0.01 (–0.7, 0.7)0.974
 TotalAge adjusted mean difference: 0.05 (95% CI –0.1, 0.2) p= 0.49
Female35–441.8 (1.7)1.5 (1.7)0.3 (0.04, 0.5)0.020
 45–542.6 (2.4)3.2 (3.0)–0.6 (–0.8, –0.5)<0.001
 55–645.8 (4.1)5.0 (3.7)0.8 (0.5, 1.0)<0.001
 65–749.1 (5.9)8.0 (4.8)1.2 (0.6, 1.7)<0.001
 TotalAge adjusted mean difference: –0.3 (95% CI –0.5, –0.2) p<0.001


The PREDICT study has assembled the largest cohort of Indian people in New Zealand with documented CVD risk profiles. The age distribution was younger for Indians than Europeans, indicating that primary health providers are following national recommendations to screen Indians 10 years earlier.1 More Indian people are likely to have diabetes with a smaller proportion likely to be smokers than Europeans. They also have a lower mean systolic blood pressure level. However, the predicted age-adjusted Framingham five-year CVD risk scores were similar in both ethnicities, with the increase in diabetes among Indians being counterbalanced by their lower smoking prevalence and lower blood pressure levels. This was an unexpected finding given their higher CVD hospitalisation and mortality rates compared with other New Zealanders.2,3

Primary care practitioners using electronic templates to gather standardised information reduced the likelihood of any differential measurement error. The NZDep Index uses aggregated population data to infer an individual's socioeconomic status and is subject to measurement error but is likely to be non-differential. Our study showed that Indian people were over-represented in areas with higher deprivation, a finding consistent with another national study.2 Our analyses only compared Indians and Europeans with completed risk assessments from consenting PHOs and, at the time of analyses, only 15% of eligible patients had been risk assessed. Therefore participants are not necessarily a representative sample of their respective population groups. Our findings cannot be considered to estimate population-based prevalences of CVD risk factors. However, this comparison was made on the assumption that there were no systematic differences in the selection process for CVD risk assessment by ethnicity except for guideline recommendations to risk assess Indian people 10 years earlier. While Indian people comprise the largest South Asian group in New Zealand, current national ethnicity coding systems preclude the opportunity to identify other South Asian ethnicities.

With these caveats, we found that the baseline risk distributions for study participants were broadly consistent with previous research. Studies in the United Kingdom have reported similar findings in South Asian populations.6–7 New Zealand studies have shown that Indian females have a considerably lower likelihood of smoking compared with European females2 and the prevalence of self-reported diabetes is higher among Asian peoples (highest in South Asians) than Europeans.3 Self-reported diabetes among Indian people has been reported as being three times that of the total New Zealand population.2 In our study, Indian participants had a lower mean systolic blood pressure than Europeans. A systematic review in the UK found that only five of 10 studies reported a higher prevalence of hypertension in South Asian males compared with Europeans.8

While the distributions of individual risk factors differed in several respects among both ethnicities, the predicted age-adjusted mean Framingham five-year risk scores were much the same. Recent national hospitalisation and mortality data suggests that South Asians living in New Zealand have a higher prevalence and incidence of CVD than European New Zealanders.9 Hence the similar Framingham risk scores suggest that the Framingham risk equation may underestimate risk in Indian people compared with Europeans. Results from the New Zealand Diabetes Cohort Study10 found that Indian participants with diabetes were at significantly higher risk of an ischaemic CVD event than Europeans (after controlling for traditional risk factors), and that the Framingham equation underestimated the observed event rate. As the PREDICT study increases in numbers and follow-up time, it will be possible to directly validate the Framingham–based prediction equations (for those with and without type 2 diabetes) and to develop new equations for Indian people. However, without changes to national ethnicity coding rules we will not be able to do this for all South Asian ethnicities.


We thank all health care providers and patients who participated. PREDICT has been supported by HRC grants 03/183 and 08/121 and was developed by the University of Auckland, Enigma Publishing Ltd, primary health care organisations, non-governmental organisations, district health boards and the Ministry of Health. PREDICT is a trademark of Enigma Publishing Ltd.