Discussion of main findings
In our study, total PA was associated with HDL cholesterol and triglyceride levels in the African-Surinamese group, but not in the European-Dutch and Hindustani-Surinamese groups. This lack of association with total PA in the European-Dutch group was particularly surprising. Total PA (from multiple domains) has been associated with more beneficial levels of HDL cholesterol and triglycerides in other study populations of European descent [17, 18]. Nevertheless, there are other patterns, for example, a lack of association of PA with HDL cholesterol levels in the European population in Newcastle upon Tyne .
The explanations for these different findings are unclear. However, it is very likely that differences in study methods may contribute to differences in the findings. For example, Fransson et al.'s study used a single-scale PA self-report measurement (with either 15 categories or 3 categories) for each domain that did not identify the components of duration, frequency, and intensity separately . Hayes et al. measured sports and exercises by duration in moderate or vigorous intensity (based on METs), while categorising walking and cycling by distance only . Therefore, intense walking and cycling might have been missed, which could possibly diminish the association in the study sample. In our study, however, the participants could report every component (duration, frequency, and intensity) for every activity, and we also found a lack of association of our total PA score with HDL levels. Alternatively, it might be that the absolute level of PA was too low to adequately model the variation in blood lipids (resulting in wider confidence limits) in this study sample, particularly in the European-Dutch and Hindustani-Surinamese ethnic groups.
Therefore, we also focussed on the contributions of the components of PA. Because we took the intensity and duration of PA into account, associations that were more similar across the ethnic groups became visible. This is in line with other multi-ethnic studies that found beneficial associations of the HDL cholesterol level with intensity of activity, although most evidence comes from studies including only sports and exercise in leisure time PA and not total PA . Triglyceride levels were not associated with intensity of activity in African Americans in the study by Monda et al. . The explanation given was the more favourable lipid profile of the African participants. We also observed a more favourable triglyceride level among the African-Surinamese in our study; however, there was also an association between intensity of activity and triglyceride level in this population. Other lifestyle factors, for instance the influence of diet on triglyceride level, could be responsible for this different association of physical activity between these groups.
The only instance where we were unable to see an association with intensity of activity was in the HDL cholesterol level of the Hindustani-Surinamese group. We believe this was more likely due to the low levels of reported PA in this group, resulting in less precise estimates and wider confidence intervals. Low levels of PA in populations of South Asian descent have been reported before . Even though we did not observe an association, there is no reason to doubt the beneficial effect of intense PA on the blood lipids of South Asians. For example, Misra et al. found a relationship between heavy activity and HDL cholesterol in Asian Indian men . In another example, a 6-week training programme for female Indian boxers showed a favourable effect on their HDL cholesterol levels .
The intensity of the activity contributed more to the association in the HDL cholesterol and triglyceride levels in our multi-ethnic study population than the duration of activity. We observed this lack of beneficial association of duration of activity with blood lipids for the whole study population (there was an indication of an opposite association in the African-Surinamese group, although it was not statistically significant in an overall test for interaction). One explanation for this lack of association is that most of the duration of activity in our study population comes from less intense activities, which have a weaker relationship with blood lipids. For example, Westerterp and Plasqui's work show that vigorous PA is often of short duration and, therefore, most of the duration of total activity comes in the form of less intense activities . The observation that duration of activity seems inversely related to HDL in African-Surinamese, might be related to this explanation. While intensity contributed more to the association with blood lipids in our study population than duration, there is evidence in randomized controlled trials that, besides an effect of intensity on blood lipids, the amount of activity (distance) and duration is positively related to the blood lipid profile [25, 26]. While the amount is closely related to duration, the duration and intensity of activity in these trials were generally greater than in our study population (for example, brisk walking or jogging a distance of 10 or 15 km three times a week).
The results of this study suggest that the inconsistent association with the total PA measure might be due to the varying composition of the PA between ethnic groups, and thus might relate to the measurement of PA. These findings are very relevant to other multi-ethnic populations living in a different context, especially when patterns of PA (in terms of duration and intensity) are markedly different between population groups. In addition to a measure of total PA or energy expenditure, separate presentations of the composition of the PA, especially intensity of activity, is needed in PA research.