These data, based on more than one million US blood donors, show a prevalence of obesity that is modestly lower than in the US general population, but still much higher than public health targets of 15 % obesity among adults(11)
. We also found associations with age, race/ethnicity and education similar to those found in NHANES. Taken together, the data support the concept that the blood centre may be a useful venue for monitoring population trends in obesity and, potentially, introducing interventions towards the maintenance of healthy BMI.
Compared with age- and sex-specific NHANES data(4)
, blood donors had relative prevalences of obesity that were modestly lower in all age and sex subgroups, except for men aged 20–59 years who had obesity prevalence comparable to NHANES. These differences are likely due to selection bias operating on donors, including selection by the blood banks for healthy individuals and self-selection for donation by individuals with higher educational achievement, which is related to lower rates of obesity(12)
The distribution of BMI in both sexes showed a strong dependence on age. Most of this effect was likely due to biological factors, namely reduced metabolic demands in older persons coupled with a continued high caloric diet in the USA. However it is also conceivable that lower BMI in the youngest age groups represents a secular trend towards reduced BMI, as previously suggested(13,14)
. Thus, recent public health educational efforts directed towards the young or changes in lifestyle may become durable as these groups age. Because of the large size and ongoing nature of the available data sets, prospectively gathered blood centre data could provide an excellent opportunity for dissecting the effects of biological ageing and secular trends in the prevalence of obesity.
Projections of future trends in obesity have substantial implications for public health policy and expenditures related to CVD and other adverse outcomes of obesity. Olshansky et al
. were the first to postulate a potential decline in US life expectancy during the 21st century due to obesity, despite gains made in the reduction of other cardiovascular risk factors(15)
. For example, gains in life expectancy related to decreased prevalence of cigarette smoking, hypertension and dietary fat intake may be counterbalanced by an increasing prevalence of obesity(1,16)
. Obesity may also be related to increased incidence of pancreatic and prostate cancer(17,18)
, decreased health-related quality of life(19)
, and increased mortality and decreased healthy survival in the elderly(3,20)
. On a positive note, some authors have suggested that the epidemic of increasing obesity from 1970 to 2000 has begun to level off in the last decade, but additional data are needed to confirm this(4,21)
We also showed strong associations between obesity and race/ethnicity, country of birth and gravidity. Donors of black race/ethnicity, especially women, were more likely to have higher BMI and obesity prevalence, as observed in NHANES(4)
. However those authors caution that differences in BMI between race/ethnic groups do not directly correlate with adiposity since muscle to fat ratios may differ by ancestry. Likewise, different race/ethnic groups may have difference risks for CVD or other adverse outcomes of obesity for any given BMI. As other authors have noted particularly among Hispanics, we observed that foreign-born donors had lower prevalence of obesity than US-born donors, consistent with the observation that immigrants have healthier diets than US-born donors(22)
. Although women with a previous pregnancy were more likely to be obese than nulliparous women, we did not see a strong relationship between obesity and higher gravidity values, consistent with a previous report based upon NHANES data(23)
Whereas obesity was more prevalent in those with some college education than in those with high school or lower and college or higher education, the multivariate analysis showed an inverse relation with educational attainment. Higher socio-economic status is generally inversely associated with obesity in high-income countries but directly associated with obesity in lower-income countries(24)
. However, data in the USA support a weakening of the inverse association between socio-economic status and obesity in recent decades, particularly among blacks(25)
. Our data also showed regional differences in obesity prevalence within the USA that are similar to those reported elsewhere, and likely represent regional differences in diet and exercise(5)
We showed marked differences in obesity prevalence according to the type of blood donation made by the donor, in order of increasing obesity prevalence: whole blood donors; platelet apheresis donors; and double red cell donors. Compared with unadjusted OR, the aOR accounting for confounding by covariates showed some reduction for platelet donors but not for double red cell donors. This indicates that selection of donors for specific donation procedures according to body weight results in enrichment of obese donors. Whereas specific weight criteria are applied to double red cell donors in order to guarantee minimum blood volume, there are not overt weight criteria for platelet donation, although donors may be selected according to previous platelet yield, which in turn may be related to blood volume and body weight. On the other hand, the higher prevalence of obesity in repeat donors was attenuated in the multivariate analysis, suggesting confounding by age or other variables as well as selection for heavier (larger blood volume) repeat donors who can better maintain iron stores with repeated phlebotomy. We were surprised to find an increased prevalence of obesity among donors with a history of receiving a blood transfusion, with aOR =1·11 (men) and 1·18 (women) after multivariable adjustment. Perhaps illnesses associated with obesity are also associated with an increased likelihood of blood transfusion, even among generally healthy blood donors. Finally, blood bankers should be aware of the high prevalence of obesity among female donors selected for automated collections, including the possibility that formulas based on height and weight may overestimate blood volume in obese donors. In order to prevent hypotensive reactions that are more frequent in donors with lower blood volume, such formulas may need to be adapted to account for adiposity v. lean body mass.
Strengths of the study include its very large population size, uniform collection of height and weight data as well as other covariates, and the ongoing nature of data collection in the blood centre setting. Limitations include the use of self-report instead of direct measurement of height and weight, because underestimation of weight and overestimation of height by respondents may have led to underestimation of BMI. A Swiss study found that self-reported height and weight underestimates BMI by 0·8 kg/m2
in men and 1·0 kg/m2
in women, and proposed an algorithm for transforming self-reported data(26)
. A similar approach could be used to standardize US self-reported blood donor data to measured height and weight in a subsample of donors. Another limitation is the exclusion of individuals with very low body weight from blood donation and hence from participation in our data. Only 2326 (<0·3 %) prospective donors were deferred for low body weight during our study period, although more persons may have self-deferred and not attempted to donate. Finally, a ‘healthy donor’ effect, namely selection bias of healthy individuals for blood donation, may operate in our data(12)
. The exclusion for low weight would imply overestimation of BMI, whereas the ‘healthy donor effect’ would tend to exclude extremes of BMI. Nevertheless our data are comparable to albeit slightly lower than NHANES BMI data, suggesting that the extent of such biases is relatively modest.