Age Groups

GBD 2019 estimated prevalence of exposure and attributable deaths, YLLs, YLDs, and DALYs for 23 age groups; males, females, and both sexes combined; and 204 countries and territories that were grouped into 21 regions and seven super-regions. GBD 2019 includes subnational analyses for Italy, Nigeria, Pakistan, the Philippines, and Poland, and 16 countries previously estimated at subnational levels (Brazil, China, Ethiopia, India, Indonesia, Iran, Japan, Kenya, Mexico, New Zealand, Norway, Russia, South Africa, Sweden, the UK, and the USA). All subnational analyses are at the first level of administrative organisation within each country except for New Zealand (by Māori ethnicity), Sweden (by Stockholm and non-Stockholm), the UK (by local government authorities), and the Philippines (by province). In this publication, we present subnational estimates for Brazil, India, Indonesia, Japan, Kenya, Mexico, Sweden, the UK, and the USA; given space constraints, these results are presented in appendix 2. For this cycle, nine countries and territories (Cook Islands, Monaco, San Marino, Nauru, Niue, Palau, Saint Kitts and Nevis, Tokelau, and Tuvalu) were added, such that the GBD location hierarchy now includes all WHO member states. These new locations were previously included in regional totals by assuming that age-specific rates were equal to the regional rates. At the most detailed level, we generated estimates for 990 locations. The GBD diseases and injuries analytical framework generated estimates for every year from 1990 to 2019.

The size of the study sample population required to reach 100 qualified participants per decile for Cam-CAN Stage 2 is expected to vary by age when accounting for exclusion and refusal, estimated population data, clinical based experience and estimates of individuals who may refuse to participate in neuroimaging. Numbers are adjusted for the proportion of the general population with exclusion criteria including MR safety contraindications (e.g. pacemakers), learning disability (living at home), cognitive impairment (Mini-Mental State Examination (MMSE) [8 (link)] score of 24 or less) and reduced response from individuals with limited longstanding illness or disability. Proportions are estimated based on data from the Office of National Statistics (ONS), the Medical Research Council Cognitive Function and Ageing Study (MRC-CFAS) [9 ] and the National Health Service (NHS) registrations. We assume that only 30% of the population will undertake the initial interview and of those who do, 40-50% will agree to take part in Stage 2 (age dependent). Numbers predicted to be needed for Stage 1 are shown in Table 2. The age group above age 88 are recruited to the same population proportion as the 78-87 decile, in order to enable cohort comparison with other population-based studies and investigation of the rare group of oldest old who are experiencing healthy ageing.Table 2

Estimated Stage 1 recruitment across the deciles to recruit 100 participants in each decile (age 18-87) for Stage 2

	Decile 1 (18-27 years)	Decile 2 (28-37 years)	Decile 3 (38-47 years)	Decile 4 (48-57 years)	Decile 5 (58-67 years)	Decile 6 (68-77 years)	Decile 7 (78-87 years)	Decile 8 (88+ years)
Contact	750	775	850	950	1250	1400	2850	1700
Interview	250	250	275	300	400	450	850	500

Estimates include numbers per decile to be contacted and interviewed.

The Cam-CAN structure provides sufficient sample size in each decile to separate age-related change from other sources of individual variation. A number of different comparisons can hypothetically be undertaken using this structure. All hypotheses are investigated at a power of 80% and α = 0.05: for linear regression, assuming the continuous data are standardised to a N(0,1) distribution, 100 per decile enables us to investigate i) a linear decline of ±0.04 across the age range; ii) a difference in linear regression slope of size ±0.06 between two risk factor groups with a prevalence of 50% (such as gender); iii) differences in the mean values of two groups (defined with 50% prevalence) of ±0.2; iv) for dichotomous outcomes with prevalence of 0.5 in one group to detect a difference of at least ±0.1. This sample is sufficiently large to be able to detect non-linear change with age, such as a change in rate of decline, and the required size to detect stability with age (to exclude a slope of up to ±0.03 per decile). Multiple hypotheses can also be undertaken, such that linear decline of slope 0.1 can be detected for 100 independent investigations protecting the type I error rate (false positives).

We first evaluated trends in the prevalence of obesity, MUO, and MHO among all study participants from 1999 to 2018. Prevalence estimates were age standardized to the 2000 US Census population, using 3 age groups (20-39, 40-59, and ≥60 years) by the direct method. To calculate the number of individuals with obesity, MUO, or MHO, we next multiplied age-standardized prevalence estimates by the total noninstitutionalized adult population for each NHANES cycle.³² Trends in MHO proportion and individual metabolic indicators among those with obesity were then evaluated overall and by age group, sex, race and ethnicity, education level, income-to-poverty ratio, home ownership, and health insurance type. Proportion estimates were age standardized to all nonpregnant adults with obesity in the 2015-2018 NHANES cycles, using the same 3 age groups. To improve the reliability and precision of weighted estimates, 2 adjacent cycles were combined in consideration of the low prevalence of MHO. Linear trends over time were evaluated using logistic regression after regressing MHO on survey cycles (modeled as a continuous independent variable). Factors associated with metabolic health among adults with obesity were further identified with logistic regression models, adjusting for age group, sex, and race and ethnicity.
The complex survey design factors for the NHANES, including sample weights, clustering, and stratification, were accounted for as specified in the NHANES statistical analysis guideline.²⁴ We used morning fasting subsample weights in all analyses to produce estimates representative of the US population. Standard errors were estimated with Taylor series linearization. Complete case analysis was applied if the missing data level for analyses was 10% or less. Several sensitivity analyses were conducted to evaluate the impact of different criteria on MHO trends. First, information on self-reported cholesterol medication use was also used to define MUO and MHO. Second, individuals with a previous diagnosis of cardiovascular disease (CVD) were regarded as having MUO, regardless of their metabolic status.^{33 (link)} Third, abdominal obesity was used as a surrogate of general obesity in the definitions of MHO and MUO. Finally, other definitions commonly used by previous studies based on MetS components,^{29 (link),30 (link)} insulin resistance,^{4 (link)} or together with inflammation^{5 (link),6 (link)} were used to define metabolic health (eTable 2 in Supplement 1).
All analyses were performed with SAS, version 9.4 (SAS Institute Inc). Two-sided P < .05 was considered statistically significant. Adjustment for multiple comparisons was not performed as in previous reports,^{1 ,34 (link)} and the results should be interpreted as exploratory due to the potential for type I error. Statistical analyses were conducted from November 2021 to August 2022.