The authors have declared that no competing interests exist.
Conceived and designed the experiments: VVK ARR AKW LL OP DGC SGW PHW CGO. Performed the experiments: VVK ARR AKW LL OP DGC SGW PHW CGO. Analyzed the data: VVK ARR. Contributed to the writing of the manuscript: VVK ARR AKW LL OP DGC SGW PHW CGO. Data collection: ARR AKW LL OP DGC SGW PHW CGO.
Adult overweight and obesity are leading global risk factors for mortality and disability, particularly in middle and high income countries.
Further elucidation of the impact of adiposity in early adult life on cardiovascular disease and diabetes requires data from long-term cohort studies with accurate information on adiposity in early adult life, middle-age and later adult life as well as long-term prospective data on cardiovascular and diabetes risk.
The British Regional Heart Study (BRHS) is a long-term prospective study of risk factors for cardiovascular disease and type 2 diabetes among middle-aged and older British men. It is based on 7735 men (78% response) born between 1919 and 1939 who were recruited in 1978–80 aged 40–59 years from a single General Practice in each of 24 British towns.
At both assessments, men provided information on medical history, use of regular medications and lifestyle behaviour (including smoking status, physical activity, alcohol intake and social class). Standing height and weight were measured with participants in light clothing without shoes; height was measured to the last complete 0.1 cm using a Harpenden stadiometer, and weight to the last 0.1 kg using regularly calibrated scales. At the 20 year assessment two resting measurements of seated blood pressure were measured in the right arm using an automated blood pressure monitor (Dinamap 1846, Critikon Corporation, Tampa, FL, USA) with an appropriately sized cuff; values were adjusted for overreading of systolic blood pressure by the instrument
With ethical committee approval, military records were sought for all study participants (including deceased participants) except those who refused consent or confirmed that they had not undertaken military service. All information obtained from military records was de-identified prior to analysis, being linked only on the basis of the unique study identifier. Military service records (electronically indexed by name, date or birth, and service number) were retrieved from the single UK repository (TNT Ltd, Swadlincote, Derbyshire) and reviewed by a trained researcher. All measures of weight and height recorded at entry to the armed services and at other times during service were recorded
Weight at 21 was obtained using armed service records or, where not available, from self-reported weight at age 21 recalled in 1996. For armed service records, weight recorded between 20 and 22 years, using the value recorded closest to 21 years, was used as a measurement of weight at 21 years without adjustments. Where measurements between 17–19 or 23–25 years (but not 20–22 years) were available, weight at 21 years was estimated from a multi-level model fitted using all weight measurements recorded between 17 and 25 years of all individual participants, adjusting for age (allowing for a quadratic relationship) and period of enlistment (1934–38, 1939–45, 1946–50, 1951–55, 1956–75). The model included a random intercept and random slope for the linear term of age to take account of clustering of measures within individual. When recalled weight in 1996 was used, this was adjusted for recall bias quantified by comparisons of measured (within the age range 20–22 years and closest to age 21) and recalled weight at 21 years in 694 individuals with data from both sources. Recall was associated with a small overestimation of weight at 21 years in men who were thinner than average (mean estimated bias across the range of measured weight was 1.5 kg with a maximum of 6.9 kg estimated at the lowest limit of the range of measured weight) (Figure S1 in
Physical and metabolic factors were summarised by quintiles of BMI-21 to examine patterns of association. Linear regression was used to assess the associations between BMI-21, BMI-50, and BMI-70 and cardiovascular and metabolic risk markers in late adulthood (adjusting for age at outcome and town as a fixed effect); percentage changes were examined for log transformed markers. The associations between BMI-70 and cardiometabolic markers were stratified by quintiles of BMI-21 to examine effect modification using a test for trend obtained from meta-regression. Patterns of high BMI over the life course were defined according to whether an individual was below (0) or on/above (1) the 75th percentile of the BMI distribution at each age period (early, middle, later adult life) and were denoted by a triplet of zeros and ones (8 trajectories in all). The regression coefficients for all trajectories were used to examine life course patterning of cardiometabolic risk, adjusting for age at outcome and town (saturated model). Other life course models including sensitive period, accumulation and exposure change models
Among 4252 men who attended for examination at both mean 50 and 70 years (78% and 77% response rate among survivors respectively), information on weight at 21 years was available for 3464 (82%) men. Among these, 1252 men had data available from military records and 2906 men had participant recall from the 1996 questionnaire; 694 men had data from both sources.
Note: Individuals may be moving across quintiles.
Characteristic | Percentile group of BMI at age 21 years | |||||||||||
All | 0%–20% | 21%–40% | 41%–60% | 61%–80% | 81%–100% | |||||||
n | Mean ± SD/% | n | Mean ± SD | n | Mean ± SD | n | Mean ± SD | n | Mean ± SD | n | Mean ± SD | |
BMI (kg/m2) | 3464 | 22.0±3.0 | 703 | 18.4±1.4 | 671 | 20.5±0.4 | 680 | 21.7±0.3 | 671 | 22.9±0.4 | 739 | 26.1±2.6 |
normal-weight (BMI<25 kg/m2) | 3043 | 87.9% | ||||||||||
overweight (25 kg/m2≤BMI<30 kg/m2) | 367 | 10.6% | ||||||||||
obese (BMI≥30 kg/m2) | 54 | 1.6% | ||||||||||
Weight (kg) | 3464 | 66.8±10.4 | 703 | 55.8±6.0 | 671 | 62.3±4.2 | 680 | 65.7±4.4 | 671 | 69.6±5.1 | 739 | 79.7±10.1 |
Height (m) | 3464 | 1.74±0.06 | 703 | 1.74±0.06 | 671 | 1.74±0.06 | 680 | 1.74±0.06 | 671 | 1.74±0.06 | 739 | 1.75±0.06 |
Age (years) | 3464 | 48.8±5.5 | 703 | 49.6±5.7 | 671 | 49.0±5.3 | 680 | 48.8±5.4 | 671 | 48.3±5.3 | 739 | 48.2±5.5 |
BMI (kg/m2) | 3464 | 25.4±3.0 | 703 | 23.7±2.9 | 671 | 24.4±2.5 | 680 | 25.1±2.4 | 671 | 25.9±2.5 | 739 | 27.6±3.0 |
normal-weight (BMI<25 kg/m2) | 1628 | 47.0% | ||||||||||
overweight (25 kg/m2≤BMI<30 kg/m2) | 1606 | 46.4% | ||||||||||
obese (BMI≥30 kg/m2) | 230 | 6.6% | ||||||||||
Weight (kg) | 3464 | 77.0±10.3 | 703 | 71.8±9.7 | 671 | 73.9±8.7 | 680 | 75.9±8.6 | 671 | 78.5±9.1 | 739 | 84.2±10.4 |
Height (m) | 3464 | 1.74±0.06 | 703 | 1.74±0.07 | 671 | 1.74±0.06 | 680 | 1.74±0.06 | 671 | 1.74±0.07 | 739 | 1.74±0.07 |
Age (years) | 3464 | 68.6±5.5 | 703 | 69.4±5.7 | 671 | 68.7±5.3 | 680 | 68.6±5.4 | 671 | 68.1±5.3 | 739 | 68.0±5.5 |
BMI (kg/m2) | 3464 | 27.0±3.7 | 703 | 25.3±3.6 | 671 | 26.1±3.3 | 680 | 26.5±3.2 | 671 | 27.4±3.2 | 739 | 29.3±3.9 |
normal-weight (BMI<25 kg/m2) | 1054 | 30.4% | ||||||||||
overweight (25 kg/m2≤BMI<30 kg/m2) | 1799 | 51.9% | ||||||||||
obese (BMI≥30 kg/m2) | 611 | 17.6% | ||||||||||
Weight (kg) | 3464 | 80.3±12.5 | 703 | 74.9±11.6 | 671 | 77.5±11.0 | 680 | 78.8±10.9 | 671 | 81.8±11.5 | 739 | 87.9±13.1 |
Height (m) | 3464 | 1.72±0.06 | 703 | 1.72±0.07 | 671 | 1.72±0.06 | 680 | 1.72±0.06 | 671 | 1.73±0.07 | 739 | 1.73±0.06 |
Total cholesterol (mmol/L) | 3287 | 6.0±1.1 | 663 | 6.0±1.1 | 635 | 6.1±1.1 | 642 | 6.0±1.0 | 642 | 6.0±1.1 | 705 | 6.0±1.1 |
LDL cholesterol (mmol/L) | 3244 | 3.9±1.0 | 654 | 3.8±1.0 | 627 | 4.0±1.0 | 631 | 3.9±0.9 | 637 | 3.9±1.0 | 695 | 3.9±1.0 |
SBP (mmHg) | 3450 | 148.8±24.2 | 702 | 148.6±25.1 | 667 | 149.1±24.2 | 679 | 148.4±24.1 | 665 | 148.8±23.6 | 737 | 148.8±24.2 |
DBP (mmHg) | 3450 | 85.1±11.1 | 702 | 84.6±11.6 | 667 | 84.8±10.8 | 679 | 84.6±11.1 | 665 | 85.9±11.0 | 737 | 85.4±11.1 |
Insulin (µΙU/mL)a | 3280 | 8.7 (5.7, 12.4) | 666 | 8.6 (5.5, 12.6) | 632 | 8.5 (5.6, 12.3) | 642 | 8.3 (5.5, 11.8) | 641 | 8.3 (5.6, 11.6) | 699 | 9.6 (6.4, 13.7) |
Glucose (mmol/L)a | 3287 | 5.8 (5.2, 6.1) | 664 | 5.8 (5.3, 6.0) | 636 | 5.8 (5.2, 6.1) | 641 | 5.8 (5.2, 6.1) | 643 | 5.8 (5.2, 6.0) | 703 | 5.9 (5.2, 6.2) |
HbA1c (%) | 3287 | 5.0±0.9 | 666 | 5.0±0.9 | 635 | 5.0±0.8 | 642 | 5.0±0.8 | 643 | 5.0±0.9 | 701 | 5.1±1.0 |
HbA1c (mmol/mol) | 3280 | 31.3±9.8 | 665 | 31.4±9.5 | 634 | 30.8±9.0 | 641 | 30.8±9.0 | 641 | 31.3±10.2 | 699 | 32.3±11.0 |
n: Number of participants (subset of individuals with BMI available at all age periods).
SD: Standard deviation.
a: Estimates correspond to geometric mean (Q1, Q3).
Quintiles of BMI at 21 years are calculated using all available data (including individuals with missing BMI at mean age 50 or 70 years).
Outcome (mean age 70 years) | BMI at mean age 21 years | BMI at mean age 50 years | BMI at mean age 70 years | ||||||
n | Coef (95% CI) | p-value | n | Coef (95% CI) | p-value | n | Coef (95% CI) | p-value | |
Total cholesterol (mmol/L) | 3287 | –0.009 (–0.022, 0.003) | 0.15 | 3287 | –0.007 (–0.020, 0.005) | 0.260 | 3287 | 0.011 (0.001, 0.021) | 0.03 |
LDL cholesterol (mmol/L) | 3244 | –0.007 (–0.018, 0.005) | 0.25 | 3244 | –0.009 (–0.020, 0.0022) | 0.115 | 3244 | 0.005 (–0.005, 0.014) | 0.32 |
SBP (mmHg) | 3450 | 0.123 (–0.149, 0.395) | 0.37 | 3450 | 0.492 (0.223, 0.761) | <0.001 | 3450 | 0.696 (0.481, 0.910) | <0.001 |
DBP (mmHg) | 3450 | 0.083 (–0.043, 0.210) | 0.20 | 3450 | 0.192 (0.067, 0.318) | 0.003 | 3450 | 0.336 (0.236, 0.436) | <0.001 |
Insulin (% difference) | 3280 | 1.525 (0.771, 2.285) | <0.001 | 3280 | 7.614 (6.862, 8.371) | <0.001 | 3280 | 8.587 (8.014, 9.164) | <0.001 |
Glucose (% difference) | 3287 | 0.362 (0.127, 0.598) | 0.003 | 3287 | 1.096 (0.863, 1.330) | <0.001 | 3287 | 0.733 (0.545, 0.922) | <0.001 |
HbA1c (%) | 3287 | 0.016 (0.006, 0.026) | 0.002 | 3287 | 0.046 (0.036, 0.056) | <0.001 | 3287 | 0.030 (0.022, 0.038) | <0.001 |
HbA1c (mmol/mol) | 3280 | 0.177 (0.067, 0.288) | 0.002 | 3280 | 0.488 (0.380, 0.597) | <0.001 | 3280 | 0.322 (0.234, 0.410) | <0.001 |
n: Number of participants (subset of individuals with BMI available at all age periods).
Coef: regression coefficient represent difference in cardiovascular risk markers for 1 kg/m2 increase in BMI. Estimates are adjusted for age at the time when the risk markers were measured, and town as a fixed effect.
Sensitive period | ||||||||||||
Outcome (mean age 70 years) | Never had high BMI | High BMI at mean age 21 years | High BMI at mean age 50 years | High BMI at mean age 70 years | ||||||||
n | K | Coef (95% CI) | n | K | Coef (95% CI) | n | K | Coef (95% CI) | n | K | Coef (95% CI) | |
Total cholesterol (mmol/L) | 1898 | 0% | reference | 868 | 44.12% | –0.07 (–0.16, 0.02) | 717 | 70.57% | –0.07 (–0.16, 0.03) | 802 | 100% | –0.01 (–0.10, 0.08) |
LDL cholesterol (mmol/L) | 1880 | 0% | reference | 858 | 43.94% | –0.08 (–0.15, 0.00) | 698 | 70.2% | –0.10 (–0.18, −0.01) | 783 | 100% | –0.06 (–0.14, 0.02) |
SBP (mmHg) | 1981 | 0% | reference | 909 | 46.09% | 1.82 (–0.06, 3.70) | 768 | 71.61% | 3.18 (1.17, 5.18) | 871 | 100% | 4.56 (2.64, 6.47) |
DBP (mmHg) | 1981 | 0% | reference | 909 | 46.09% | 0.99 (0.11, 1.87) | 768 | 71.61% | 1.63 (0.70, 2.57) | 871 | 100% | 2.03 (1.14, 2.92) |
Insulin (% difference) | 1896 | 0% | reference | 862 | 44.43% | 24.45 (18.42, 30.78) | 712 | 70.93% | 55.04 (47.01, 63.51) | 803 | 100% | 68.66 (60.49, 77.26) |
Glucose (% difference) | 1898 | 0% | reference | 869 | 44.07% | 3.02 (1.37, 4.69) | 713 | 70.41% | 7.35 (5.52, 9.21) | 801 | 100% | 5.68 (3.94, 7.44) |
HbA1c (%) | 1901 | 0% | reference | 866 | 44.46% | 0.16 (0.09, 0.23) | 713 | 70.97% | 0.33 (0.26, 0.40) | 804 | 100% | 0.27 (0.20, 0.34) |
HbA1c (mmol/mol) | 1896 | 100% | reference | 864 | 44.44% | 1.76 (1.00, 2.51) | 712 | 70.93% | 3.58 (2.77, 4.39) | 803 | 100% | 2.92 (2.15, 3.70) |
n: Number of participants (subset of individuals with BMI available at all age periods).
K: percentage of individuals who had high BMI (BMI in the top 25% of the distribution) also at mean age 70 years.
Coef: regression coefficient for the effect of high BMI (BMI in the top 25% of the distribution) at each sensitive period (mean age 21, 50 or 70 years) as compared with never having had high BMI. Estimates are adjusted for age at the time when the risk markers were measured, and town as fixed effect.
For each outcome, results are obtained from 3 models fitted separately (one for each period: mean age 21, 50 and 70 years).
BMI at mean age 70 years (kg/m2) | Total cholesterol | LDL cholesterol | SBP | DBP | Insulin | Glucose | HbA1c | HbA1c | ||||||||||
(mmol/L) | (mmHg) | (mmHg) | (mmHg) | (% difference) | (% difference) | (%) | (mmol/mol) | |||||||||||
BMI trajectory | n | mean (SD) | n | Coef (95% CI) | n | Coef (95% CI) | n | Coef (95% CI) | n | Coef (95% CI) | n | Coef (95% CI) | n | Coef (95% CI) | n | Coef (95% CI) | n | Coef (95% CI) |
0–0-0 | 1988 | 25.0 (2.3) | 1898 | 0 (ref. group) | 1880 | 0 (ref. group) | 1981 | 0 (ref. group) | 1981 | 0 (ref. group) | 1896 | 0 (ref. group) | 1898 | 0 (ref. group) | 1901 | 0 (ref. group) | 1896 | 0 (ref. group) |
1–0-0 | 383 | 26.1 (2.0) | 376 | –0.07 (–0.19, 0.05) | 373 | –0.05 (–0.15, 0.06) | 380 | 0.84 (–1.78, 3.45) | 380 | 0.35 (–0.87, 1.57) | 374 | –4.68 (–10.77, 1.83) | 377 | –0.410 (–2.57, 1.79) | 375 | 0.03 (–0.06, 0.13) | 374 | 0.36 (–0.67, 1.39) |
0–1-0 | 109 | 27.2 (1.6) | 102 | –0.13 (–0.34, 0.08) | 100 | –0.16 (–0.35, 0.04) | 108 | 0.34 (–4.29, 4.97) | 108 | 1.36 (–0.79, 3.52) | 102 | 21.28 (7.67, 36.60) | 102 | 6.75 (2.61, 11.06) | 101 | 0.35 (0.18, 0.52) | 101 | 3.73 (1.86, 5.59) |
1–1-0 | 111 | 27.5 (1.2) | 109 | –0.20 (–0.41, 0.002) | 108 | –0.17 (–0.35, 0.02) | 110 | –0.67 (–5.25, 3.90) | 110 | –0.43 (–2.56, 1.70) | 105 | 18.34 (5.27, 33.02) | 109 | 4.52 (0.60, 8.60) | 106 | 0.21 (0.04, 0.38) | 106 | 2.25 (0.43, 4.06) |
0–0-1 | 219 | 30.8 (1.9) | 202 | 0.00 (–0.15, 0.16) | 201 | –0.05 (–0.19, 0.09) | 218 | 5.03 (1.69, 8.37) | 218 | 1.49 (–0.06, 3.05) | 203 | 69.01 (55.04, 84.24) | 202 | 2.27 (–0.63, 5.26) | 201 | 0.16 (0.03, 0.28) | 201 | 1.64 (0.28, 2.99) |
1–0-1 | 103 | 30.7 (2.0) | 94 | 0.06 (–0.16, 0.28) | 92 | –0.05 (–0.25, 0.15) | 103 | 3.73 (–1.01, 8.47) | 103 | 2.78 (0.57, 4.99) | 95 | 48.90 (31.63, 68.43) | 97 | 0.60 (–3.40, 4.77) | 97 | 0.09 (–0.09, 0.26) | 97 | 0.85 (–1.05, 2.76) |
0–1-1 | 235 | 32.0 (2.8) | 217 | 0.019 (–0.13, 0.17) | 205 | –0.04 (–0.18, 0.10) | 234 | 6.12 (2.89, 9.36) | 234 | 2.63 (1.12, 4.14) | 217 | 74.94 (60.86, 90.24) | 216 | 8.12 (5.14, 11.18) | 218 | 0.37 (0.25, 0.49) | 218 | 3.96 (2.65, 5.26) |
1–1-1 | 316 | 32.8 (3.0) | 289 | –0.05 (–0.19, 0.08) | 285 | –0.09 (–0.21, 0.03) | 316 | 3.32 (0.48, 6.16) | 316 | 1.71 (0.38, 3.03) | 288 | 70.58 (58.39, 83.70) | 286 | 8.09 (5.44, 10.79) | 288 | 0.34 (0.23, 0.44) | 287 | 3.73 (2.57, 4.89) |
n: Number of participants (subset of individuals with BMI available at all age periods). For BMI at mean 70 years this includes all available data.
BMI trajectories: Each triplet corresponds to a different trajectory of high BMI at mean age 21, 50 and 70 years; with 0 and 1 denoting BMI below and above the 75th percentile of the BMI distribution, respectively. For example, (0–0-0) denoted low BMI at all age periods, whilst (0–0-1) signified high BMI at mean age 70 years only.
Coef: Estimates are differences in risk marker from BMI trajectory 0–0-0. Estimates are adjusted for age at the time when the risk markers were measured, and town as fixed effect.
Compared with the saturated model (8 BMI trajectories presented in
Additional adjustment for smoking in all models made little difference to the results. The exclusion of participants on lipid lowering medication, blood pressure lowering medication or blood glucose lowering medication did not affect the results presented in
Although adiposity in middle-age is an established risk factor for cardiovascular disease and type 2 diabetes,
Our observation that BMI in later life was a considerably more important determinant of both cardiovascular and metabolic risk than BMI in early life is in agreement with an overwhelming body of evidence showing that adiposity and weight gain in middle-age show strong, graded associations with cardiovascular disease, type 2 diabetes and their precursors in later life.
The present study found no evidence that early BMI related to blood pressure and blood cholesterol levels. Previous analyses examining the associations of BMI at different ages on blood pressure and blood cholesterol at 45 years in the 1958 British birth cohort (NCDS)
The investigation is based on a geographically and socially representative cohort study with high response rates and exceptionally high follow-up rates. This study has used a novel approach in extending an established cohort of middle aged men by using data on BMI from military service records as a source of information in early adulthood. Among men who survived to the 60–79 year re-examination, it was possible using the combination of two sources of information, military health records and participant recall, to obtain information on BMI at 21 years for a high proportion of participants (82%); information from military records was documented at the time of measurement and recalled data were substantially consistent. The availability of data from both sources in 694 men allowed adjustment for a small bias in recalled weight and validated the use of recalled weight. Although the analyses were based on slightly fewer than half of participants in the original cohort, inevitably being based on a healthy survivor group, baseline risk factor differences in men who were or were not included in the present analyses did not differ markedly (data not presented). Hence, underestimation of potential associations between early BMI and later cardiometabolic risk because of earlier mortality amongst participants at greatest risk is likely to be limited. A further concern is the extended time interval between BMI at 21 years and cardiometabolic outcomes about 50 years later, by which point the development of disease could make the detection of associations difficult. However, the results for blood lipids, blood pressure and insulin and glycaemia markers were not materially affected by exclusion of study participants receiving medications likely to have affected these outcome measures. BMI is only one of many surrogate markers of adiposity. Unfortunately, other measures of adiposity such as waist circumference, waist to hip ratio and percentage of fat mass were only available for the 60–79 year examination and their use would have undermined a key strength of the current analysis, that the same marker of adiposity was used for all age groups. Moreover, BMI is highly correlated with fat mass both in this population (at 70 years, r = 0.70, p<0.001). Although BMI is a robust measure of overall adiposity, its association with body fatness differs with age; in later life changes in muscle mass as well as fatness are important influences on BMI.
The results suggest that BMI in later life is the dominant influence on risk markers both for CHD and for type 2 diabetes. Early adult BMI level may be associated with later type 2 diabetes risk, but not with markers of CHD risk. However, it is not possible to establish whether this association reflects the positive correlation between BMI in early adult life and BMI in later life, or whether it reflects a more direct association between early BMI and diabetes risk, possibly based on the cumulative effects of BMI on insulin resistance and pancreatic beta cell function over an extended period from early adult life. This uncertainty partly reflects the limitations of the present cohort for investigating this issue, given the low levels of adiposity and the low prevalence of overweight/obesity in early adult life observed in the present cohort. However, the associations between adiposity, diabetes and CHD in the present cohort (with a lesser burden of adiposity in early adult life than that in subsequent cohorts) could act as a valuable reference point for investigating the health effects of adiposity in subsequent birth cohorts, in which adiposity in early adult life became more marked. While findings from this study are generalizable to elderly white males, relevance to women of a similar age (who show different patterns and trends in body composition, e.g. greater losses in muscle mass with age), and other ethnic groups remains to be established. This work will be carried on when cohort studies with individuals with higher levels of adiposity in early life will become available.
(ZIP)