We found a large improvement in metabolic health (defined as having a normal blood pressure, normal cholesterol and no known diabetes) between 1986 and 2009 in all weight categories and in both men and women in the two northernmost counties of Sweden. However, women were generally healthier than men. The increase in metabolic health was attenuated after the turn of the millennium, and the trend was reversed after 2004 for normal- and overweight women and obese and overweight men. Notably we have not found any previous study reporting long-term trends in metabolic health in a well-defined population using a standardized methodology. As recently reviewed [24], there is no consensus on the definition of metabolic health among the obese. As our focus was to explore the impact of obesity on CVD risk we choose to rather use the well- validated risk factors systolic blood pressure, total cholesterol and absence of diabetes. A comparison of prevalence data with other studies is not possible as the 27 studies reviewed used 30 definitions of metabolic health and prevalence ranged between 6% and 75%, with strong age dependence.
The MONICA and NHANES studies [2,7] have shown decreasing CVD risk factors over many years, which are the underlying forces that drive the increase in metabolic health. Could the decreasing metabolic health from 2000 represent a new development and be related to changing diets such as the “low carb, high fat” diet which has received a great deal of attention? Thus, a long-term and very promising trend towards improved metabolic health among the obese seems to have been broken. The recently finished 2014 MONICA survey will possibly provide some answers.
In 2009, CVD risk factors still accumulated in overweight and obese subjects, more so in men than in women. This indicates that while a larger proportion of these subjects now are metabolically healthy, it is still detrimental for one’s health to be overweight and obese, supporting finds from the Framingham and NHANES studies [2,12,25]. However, the larger improvement among the obese may help to explain the paradox of decreasing CVD while obesity increases.
Abdominal obesity was associated with a lower prevalence of metabolic health and higher prevalence of risk factors for CVD. Despite the fact that the subjects with abdominal obesity more than doubled their metabolic health over the 23-year observation period, from 6.4% to 17.3%, it was still half as common for them to be healthy, compared with those without abdominal obesity. Those without abdominal obesity were as metabolically healthy as the normal-weight individuals, and those with abdominal obesity had a similar prevalence of metabolic health as the overweight or obese individuals. As been reported in NHANES and other studies [26-28], abdominal obesity is a better predictor for unhealthiness, but we could not discern any significant difference in clustering of the three risk factors, hypertension, cholesterol and diabetes, between the methods of measuring obesity.
Interestingly, in 2009 the obese women had a greater prevalence of metabolic health than the overweight women, contrary to the expected pattern seen in men. This result may be driven by the higher cholesterol among overweight women compared with the obese, as cholesterol below 5.0 mmol/l was a criterion for metabolic health. However, we have also previously reported that between 2004 and 2009, waist circumference decreased and hip circumference increased among women in northern Sweden [3]. It is noteworthy that a protective effect of larger hip circumference adds considerably to the predictive value of waist circumference on incidence and mortality of CVD as pointed out in a recent systematic review [29].
In 2009, those subjects with metabolic health had a mean age roughly 10 years younger than those without metabolic health. However, even among the youngest, less than 60% had metabolic health. In subjects aged 65 years or more, prevalence of metabolic health slightly increased compared to those 55–64 years of age. The explanation for this is unclear. Perhaps retired persons have more time for exercise and adopt a healthier lifestyle today and ignore the latest trends such as “the low carb, high fat diet”. It is also possible that the health care system more actively diagnoses and treats hypertension and hyperlipidemia in the elderly. Younger individuals may be encouraged to change their lifestyle in lieu of treatment and may not succeed or have little contact with their health care center and therefore risk factors are not assessed.
Smoking was not associated with poorer metabolic health. However, other studies have shown smoking and obesity to have a synergistic effect on cardiovascular risk [25]. With higher education it was more likely for an individual to be of normal weight and metabolically healthy, as is supported by two Swedish studies [9,30], which found that a higher education makes one less likely to be obese and thus more likely to be metabolically healthy. However, adjustment for age and sex abolished the difference between education groups, which supports the idea that life style and primary prevention in Sweden may not bias the socially disadvantaged as much as in other countries.
Subjects with regular physical exercise more often had metabolic health, although less exercise did not explain less metabolic health in obesity or overweight. In a recent study [18] fitness was associated with metabolic health in the obese. Furthermore, fitness was shown to reduce cardiovascular mortality for overweight and obese individuals [21]. Since we had no measure of fitness, we used the self-reported physical activity as a proxy, but still our findings corroborate those fitness studies.
Strength and limitations
Self-reported weight and height are prone to bias [31], and while questionnaires can provide fairly valid estimates of known diabetes, parameters such as blood pressure, glucose, waist circumference and cholesterol need to be measured. Thus, a valid description and analysis of cardiovascular risk factors, diabetes and their relationship with obesity, on a population level must be based on a physical examination of a random sample, not only on postal questionnaires, which are a common instrument in public health research.
In the Northern Sweden MONICA study, a strict and uniform methodology has been used throughout the whole time period, from 1986 to 2009. Newer methods of analysis were adjusted after re-running older samples, and all anthropometric measurements were performed by trained staff using similar equipment and protocols. This provides a wealth of highly comparable data from an extended time period. Both internal and external validity is high, and it is possible to take common confounders such as socioeconomic status into consideration.
The major limitation of this study is a declining participation rate over the study period. This is most pronounced among the younger population and could present a problem with selection bias. Telephone interviews with the majority of the nonparticipants in the first three surveys (1986, 1990, 1994) showed that they were more likely to smoke (despite similar levels of education) and less likely to be obese or hypertensive than the participants. For the 2009 survey, nonparticipants were younger with lower education and a higher prevalence of diabetes and regular smoking. This may lead to the 2009 data painting an overly optimistic picture, as a higher prevalence of diabetes and smoking should lower the amount of metabolically healthy individuals.
Our definition of metabolic health is arbitrary but based on the most recent European guidelines for cardiovascular prevention, decided upon by all the relevant scientific organizations and systematic reviews [23]. We did not have data on triglyceride or HDL-cholesterol levels, which could have helped to further refine the concept of metabolic health but perhaps not adding much to estimating CVD risk. The variables used were those that form the basis for the cardiovascular risk score (SCORE) proposed by the European Society of Cardiology.
In the definition of metabolic health, treatment for the included risk factors was not taken into account. This could be called into question as treatment of these risk factors probably does not remove all the associated cardiovascular risk. Therefore, we performed an ancillary analysis by classifying treated subjects as not metabolically healthy even if their blood pressure and cholesterol levels were normal. The same general pattern in associations persisted although the absolute levels were lower, most notably among the obese, as expected. This sensitivity analysis strengthens our findings. The dichotomization of reported physical activity may also be a too coarse and blunt instrument to measure and explore a complicated life style such as physical activity.