Health information regarding such relationships is often obtained through the use of surveys. Population surveys are a well-recognised, and much utilised method to assess the prevalence of diseases as well as obtain other health-related information (2–4). One of the more recognised survey measures are the National Health and Nutrition Examination Surveys (NHANES) (4, 5), which have incorporated subjective survey data obtained from interviews, along with additional data derived from objective clinical assessment and laboratory data. Patient-reported surveys have the advantage of obtaining health data from large sample sizes and providing access to data that may otherwise be difficult to obtain.

In this study, data from two national surveys were evaluated to determine the relationship between different BMI categories and the prevalence of diabetes mellitus, hypertension and dyslipidaemia. The Study to Help Improve Early evaluation and management of risk factors Leading to Diabetes (SHIELD) was a self-reported survey (with no clinical or laboratory evaluation) conducted in 2004 that assessed the association of different BMI categories with diabetes mellitus, hypertension and dyslipidaemia. The comparator survey was the NHANES 1999–2002, which obtained data through interviewer-administered surveys, as well as clinical evaluations and laboratory assessments (4, 5).

The BMI distributions in SHIELD and NHANES were remarkably similar (Figure 1), in large part likely because of the fact that both the SHIELD and NHANES responses were weighted to match the US adult population, and because other epidemiologic studies have demonstrated that self-reported height and weight accurately correlated with measured height and weight (16, 17). Similarly, both SHIELD and NHANES consistently demonstrated an increase in prevalence of type 2 diabetes mellitus and hypertension with increasing BMI, with reported percentage rates across various BMI ranges that were also remarkably similar (Figures 2 and ​and33).

Additionally, both SHIELD and NHANES demonstrated gradual increases in dyslipidaemia until the BMI reached above 30 kg/m² (Figure 4). Beyond this, the prevalence peaked, and in fact estimates began to decline in NHANES. The prevalence of dyslipidaemia in NHANES was higher at each cut-off point when compared with SHIELD. This is likely related to the fact that the definition of dyslipidaemia included not only history, but also four laboratory values (TC, TG, LDL-C and HDL-C levels). An abnormality of any of these individual variables would have been recorded as ‘dyslipidaemia’. Given that NHANES included laboratory testing and SHIELD did not, it is thus not surprising that the prevalence of dyslipidaemia was reportedly higher in NHANES. Had dyslipidaemia been defined as one variable, such as an increase in LDL-C level only, then the prevalence of dyslipidaemia in NHANES would likely have been less, and the differences in the reported prevalence of dyslipidaemia between SHIELD and NHANES would have been closer. Conversely, it should be noted that the lipid level values that were chosen to define ‘dyslipidaemia’ in this study were conservative. Had more aggressive cut-off levels been used to define ‘dyslipidaemia,’ such as TC ≥ 200, TG ≥ 150, LDL-C ≥ 100 or HDL-C < 60 mg/dl, instead of the definition of dyslipidaemia used in this analysis (TC ≥ 240, TG > 200, LDL-C ≥ 160 or HDL-C < 40 mg/dl), then the diagnosis of dyslipidaemia in NHANES would have been greater, and the differences between the prevalence of dyslipidaemia in SHIELD and NHANES would have been even greater. Hence, the degree of correlation of ‘dyslipidaemia’ in a self-reported survey (such as SHIELD) compared with that of an objective survey that includes laboratory assessment (such as NHANES), and that is performed on a wide spectrum of participants (without regard to their CHD risk) is thus largely dependent upon how the dyslipidaemia is defined.

With regard to the analysis of patients with diabetes mellitus, hypertension or dyslipidaemia, whether it was data collected through a self-reported survey only (such as SHIELD) or through a more detailed evaluation (NHANES), 75% or more of patients with each of these individual metabolic diseases (often thought to be ‘obesity related’) were overweight or obese, while about 10–25% were not overweight. In fact, some prevalence of metabolic diseases was reported at all BMI levels. Collectively, the findings presented here document that, while often directly related, not all overweight or obese patients have diabetes mellitus, hypertension or dyslipidaemia, and that not all patients with these metabolic diseases are overweight or obese. This simple message has profound implications as to the pathophysiologic relationship between fat and metabolic disease (18–20), such as whether, from the standpoint of excessive fat-related metabolic diseases, it is best to focus on fat mass (adiposity) alone (Table 1), or whether a focus on the pathogenic potential of adipose tissue (adiposopathy) might also be warranted (20–27).

In other words, whether or not weight gain may cause or worsen metabolic disease (25) and whether or not weight loss may improve metabolic disease (26) are very much dependent upon the effects on the pathogenic potential of adipose tissue. For example, positive caloric balance is most likely to result in metabolic disease when accompanied by: (i) impaired adipogenesis, which limits energy storage potential, resulting in excessive adipocyte hypertrophy which adversely affects adipocyte/adipose tissue dysfunction; (ii) accumulation and hypertrophy of visceral fat, hypertrophy of peripheral fat and increases in intra-organ fat (such as in the liver, muscle or pancreas), which result in adverse metabolic and immunologic consequences; (iii) impaired nutrient metabolism such as a net increase in free fatty acids, which is lipotoxic to body organs such as muscle, liver and pancreas; (iv) adipocyte and adipose tissue dysfunction which results in adverse metabolic consequences, because adipose tissue is an active endocrine organ (v) adipocyte and adipose tissue dysfunction which results in adverse immunological consequences, because adipose tissue is an active immune organ, and (vi) disruption of optimal interorgan ‘cross-talk’ of adipose tissue with other body organs, because metabolic diseases associated with positive caloric balance are most often caused by a pathologic partnership between the dysfunction and/or limitations of adipose tissue and the dysfunction and/or limitations of other body organs (25–27). Adiposopathy is a term used to describe pathogenic adipose tissue whose adverse clinical consequences may be promoted and exacerbated by adipocyte hypertrophy, visceral adipose tissue accumulation, and sedentary lifestyle in genetically and environmentally susceptible patients, and which represents an underlying, root physiological process leading to metabolic diseases such as type 2 diabetes mellitus, hypertension and dyslipidaemia. The results of this survey study demonstrate that while generally and directly associated with one another, the relationship between BMI and metabolic disease is not an absolute one, and further lends support to the adipocentric paradigm wherein pathologenic adipose tissue (adiposopathy) is a more rational treatment target than BMI (adiposity) alone (23).

With regard to the survey itself, the SHIELD study represents the largest such initiative ever taken. However, consumer panel surveys such as SHIELD do have limitations. First of all, SHIELD relied only on self-reporting of medical data without clinical or laboratory confirmation. Furthermore, only a small percentage (5–8%) of consumers initially invited to participate in the NFO panel (the step prior to mailing of the screener questionnaire) elect to do so, leading to the possibility of bias because of self-selection. Also, household panels also tend to under-represent the very wealthy and very poor segments of the population, and do not include military or institutionalised individuals (6, 28). However, NFO survey response rates are generally high (60–75%) and the demography of non-responders is known and can be controlled for in analyses. Another potential confounder includes the potential for misreporting of parameters such as height and weight in a self-reported survey.

Nonetheless, this report demonstrates that a self-reported survey can often acquire data that reasonably approximates surveys that also include clinical and laboratory evaluations. This is important because population assessments of the frequency of the associations of obesity with metabolic diseases such as diabetes mellitus, hypertension and dyslipidaemia have great epidemiological value, but are impaired by the logistical difficulties in obtaining reasonable and reliable data to make these assessments. For the SHIELD screener survey, a large number of questionnaires were sent, with a high return rate (64%), providing a sample that is generally representative of the overall US population. Thus, the SHIELD survey appeared to be a relatively cost-effective method to collect data on many aspects of the relationship of self-reported data and metabolic diseases.

Another potential utility of a self-reported survey is that the use of a volunteer panel that is accustomed to completing surveys also allows for the collection of much data that are otherwise difficult to collect (e.g. quality-of-life data for those with diabetes mellitus). Furthermore, longitudinal surveys are more easily obtained. For example, subsamples of the SHIELD screener respondents are currently under way, using a longer, more detailed survey assessing individual health status, health knowledge, and attitudes as well as current behaviours and treatments. Annual follow-up assessments are planned over the next 4 years, which will allow further exploration of relationships between these variables and metabolic diseases.