Abstract

Change and fluctuation in body mass index (BMI; weight (kg)/height (m)2) may be associated differently with coronary artery calcification (CAC) than with carotid artery intima-media thickness (IMT). The authors analyzed data on 2,243 participants in the Coronary Artery Risk Development in Young Adults (CARDIA) Study, initially aged 18–30 years, who were examined every 2–5 years over a 20-year period (1985–2006). BMI at year 0 was associated positively and linearly with CAC at year 20; however, the association of BMI with year 20 CAC became progressively U-shaped in subsequent examinations (years 10, 15, and 20). To understand the deepening U shape, the authors modeled year 20 BMI and its history using 3 indices: year 0 BMI, linear slope of BMI during 20 years, and BMI fluctuation during 20 years. In models including these 3 terms, year 0 BMI was associated positively with CAC, as was BMI fluctuation. However, adjusted odds ratios across quintiles of BMI slope (vs. the lowest quintile) were 0.7, 0.4, 0.5, and 0.4 (Ptrend < 0.01), suggesting higher risk of CAC with weight loss, plateauing after moderate weight gain. In contrast, IMT was associated positively with BMI at all examinations and with 20-year BMI slope and was unassociated with BMI fluctuation. Surprisingly, CAC risk was higher with BMI loss and lower with BMI gain, whereas associations with IMT were as expected.

Obesity has shown different associations with coronary artery calcification (CAC) than with carotid artery intima-media thickness (IMT), although both are commonly used subclinical markers of atherosclerosis. IMT has been consistently and strongly associated with various indices of obesity, including body mass index (BMI) and waist circumference (1–4). However, the adiposity measures have not been consistently related to CAC (5–12). Different adjustments for confounders have not explained this inconsistency.

Repeated measurement of weight over time allows us to study both the effects of baseline weight and the effects of changes in weight during the observation period. In the current worldwide obesity epidemic, weight change, including fluctuation, is a common phenomenon. Therefore, in addition to baseline body weight, change and fluctuation in body weight (or equally in BMI, if adjusted for height) may also be important in determining the risk of IMT or CAC. Also, the inconsistent associations of obesity with CAC and IMT may be explained by different associations of change and fluctuation in body weight with CAC or IMT.

We used data from a cohort followed longitudinally over 20 years (1985–2006), the Coronary Artery Risk Development in Young Adults (CARDIA) Study cohort, to examine the associations of weight change and weight fluctuation with CAC and carotid artery IMT at year 20. In a published article on BMI and CAC in the CARDIA Study (8), only baseline BMI, not BMI at year 15, was positively and linearly associated with CAC at year 15. In the current study, we examined associations with both CAC and IMT by modeling year 20 BMI and its history using 3 indices: BMI at year 0, linear slope of BMI during 20 years, and BMI fluctuation around the linear slope during 20 years. Additionally, we examined associations between 3 indices of year 15 BMI and its history and the progression of CAC during 5 years from year 15 to year 20. This analytic approach has been used when repeated measurements of body weight during follow-up were available (13–15).

MATERIALS AND METHODS

Study population

In CARDIA, the evolution of cardiovascular disease risk was studied in participants who were initially aged 18–30 years (16). In brief, 5,115 African-American and white participants were recruited at baseline in 1985–1986 (year 0) in 4 US cities (Birmingham, Alabama; Chicago, Illinois; Minneapolis, Minnesota; and Oakland, California). Follow-up examinations were completed at years 2, 5, 7, 10, 15, and 20, with retention of 91%, 86%, 81%, 79%, 74%, and 72% of survivors, respectively. Institutional review board approval and informed consent were obtained at each site at every examination.

CAC was assessed at years 15 and 20 and IMT only at year 20; all of these measures were available for 2,288 participants. We replaced BMI with a missing value in females who were pregnant when examined, leading to 118 missing values in those specific examination years only. We then excluded 45 participants with missing data on BMI for that reason or any other reason at year 0, at year 20, or at 3 or more examinations among years 2, 5, 7, 10, and 15 (so that all participants had at least the initial measurement, the terminal measurement, and 5 total BMI measurements), leaving a final sample size of 2,243. Exclusion of all information from women who were ever pregnant during any examination did not change results.

Measurements

A common protocol and quality control procedures were used at all examinations. Participants were asked to fast for at least 12 hours and to avoid smoking and heavy physical activity for 2 hours before examination. Self-reported current smoking status was noted. Habitual physical activity was measured using the CARDIA Physical Activity History questionnaire (17). Body weight was measured in light clothing (to the nearest 0.2 pound (0.1 kg)), body height was measured without shoes (to the nearest 0.5 cm), and BMI was calculated as weight (kg) divided by height squared (m2). Participants were twice asked about intentionality of weight loss or gain during follow-up (at year 7 for the period between years 5 and 7 and again at year 10 for the period between years 7 and 10). Concentrations of plasma total cholesterol, high density lipoprotein cholesterol, and triglycerides were determined enzymatically by Northwest Lipids Research Laboratory (Seattle, Washington) at all examinations. The level of low density lipoprotein cholesterol was derived by means of the Friedewald equation. Serum glucose concentration was measured at year 0, using the hexokinase ultraviolet method, by American Bio-Science Laboratories (Van Nuys, California) and at years 7, 10, and 15, using hexokinase coupled to glucose-6-phosphate dehydrogenase, by Linco Research, Inc. (St. Louis, Missouri).

Two computed tomography scans were obtained at each of year 15 and 20 using electron-beam computed tomography (Imatron C-150; GE Medical Systems, Milwaukee, Wisconsin (for the Chicago and Oakland centers)) or multidetector computed tomography scanners (GE Lightspeed, GE Medical Systems (for the Birmingham center) or Volume Zoom, Siemens, Erlangen, Germany (for the Minneapolis center)) (18). For each scan, 40 consecutive images from the root of the aorta to the apex of the heart were obtained. Participants remained supine between scans taken 1–2 minutes apart. Each scan was read centrally by a trained reader. The reader identified a region of interest for each potential focus of CAC, defined as 4 or more adjacent pixels (1.87 mm2) with a computed tomography number greater than 130 Hounsfield units (field of view = 35 cm). Agatston scores were adjusted for between-center differences using a standard calcium phantom scanned underneath each participant, and scores were summed across the 4 major coronary arteries. Scans were adjudicated side-by-side within-year, if 1 scan was positive and the other was zero, or between years 15 and 20, if scans were widely discrepant or Agatston score decreased. In our data, there was greater noise in the computed tomography image in participants with higher BMIs, which led to falsely positive Agatston scores in the initial reading; these false-positive scores were reset to zero after rereading.

The thicknesses of the intima media layer of the common carotid artery at the carotid bulb and in the internal carotid artery were determined from images obtained by high-resolution B-mode ultrasonography at year 20 (Logiq 700 ultrasound machine; GE Medical Systems, Waukesha, Wisconsin). Calculation of IMT was performed at the CARDIA ultrasound reading center at Tufts-New England Medical Center (Boston, Massachusetts), using the maximum across the mean values in the near and far walls on the left and right sides, including plaque as part of the wall. Pearson correlation coefficients based on 58 replicate readings were 0.86 for the common carotid artery, 0.72 for the bulb, and 0.88 for the internal carotid artery.

Statistical analysis

First, we calculated results based on cross-sectional analyses of CAC or IMT with BMI, all at year 20, and then calculated results for associations of year 20 CAC or IMT with BMI at year 0 and with BMI at year 10. Next, we studied 3 separate indices of year 20 BMI and its history: BMI at year 0, linear slope of BMI from year 0 to year 20, and BMI fluctuation around the BMI trend line during the 20-year period. Slope of and fluctuation in BMI from year 0 to year 20 were estimated using a simple linear regression model, estimated separately for each of the 2,243 participants, in which the dependent variable was the participant's BMI and examination year was the independent variable. The slope coefficient of this model was used as an index representing the linear trend of a participant's BMI change in terms of direction and magnitude. The root-mean-square error for each subject (the residual variability around the overall time trend of BMI) was used to represent the magnitude of BMI fluctuation. Since height is essentially constant, negative BMI slope implies weight loss, positive BMI slope implies weight gain, and higher BMI fluctuation implies greater weight fluctuation. We sometimes refer to “weight” instead of BMI, recognizing that “weight change” is actually change in weight adjusted for height squared.

In this paper, we present results from analyses in which high CAC and high IMT were defined on the basis of Agatston score >20, the 90th percentile, and IMT >90th percentile value. In fact, when we used several cutoff points to represent high CAC (Agatston score >0, >10, >20, and >40) and high IMT (IMT >70th, >80th, and >90th percentile), the associations became progressively stronger the higher the cutoff point. Detailed results based on all definitions of CAC and IMT are presented in Web Tables 13, which are posted on the Journal’s Web site (http://aje.oxfordjournals.org/). Results based on continuous values of CAC and IMT were similar to those based on dichotomous values for these outcomes. Although there were 2 IMT outcomes, common carotid artery IMT and carotid bulb/internal carotid IMT, most significant associations were observed only with common carotid artery IMT. Therefore, we present results for common carotid artery IMT.

BMI at baseline, linear slope of BMI, and fluctuation in BMI were each divided into quintiles. We also present results obtained after further categorization of the lowest quintile of linear slope in BMI into weight loss and slight weight gain. In additional analyses, we excluded weight losers and studied only weight gainers. Adjustments were made for the year 20 covariates age, sex, race/ethnicity, study center, cigarette smoking, and physical activity. In additional models, we considered clinical cardiovascular disease risk factors as covariates, namely systolic blood pressure and levels of fasting glucose, low density lipoprotein cholesterol, high density lipoprotein cholesterol, and triglycerides at the year 20 examination, recognizing that these variables may be explanatory of findings related to BMI.

Because CAC was assessed at both year 15 and year 20 but IMT was assessed only at year 20, we could only examine progression for CAC. We examined increase in Agatston score >20 from year 15 to year 20 by conducting parallel analysis with the 3 components of BMI at year 15 included as independent variables and adjusting for year 15 covariates. We used SAS statistical software (version 9; SAS Institute Inc., Cary, North Carolina) for the analyses.

RESULTS

General characteristics

The 2,243 participants had a mean age of 45.3 years (standard deviation (SD), 3.5; range, 37–52) at year 20. Mean BMI increased monotonically across examinations from year 0 (24.2 (SD, 4.4)) to year 20 (29.0 (SD, 6.2)). The proportion of subjects with weight loss during 20 years was only 9.2% among all subjects (46.2% among subjects belonging to the lowest quintile of BMI slope). Table 1 shows several characteristics at year 20 according to the 3 different indices of year 20 BMI and its history: year 0 BMI, linear slope of BMI from year 0 to year 20, and BMI fluctuation during the 20-year period. Women and African Americans experienced both higher BMI gain and higher BMI fluctuation during the 20 years. Less weight gain and greater BMI fluctuation were associated with current smoking. On the other hand, more physically active persons had less weight gain and less BMI fluctuation.

Table 1.

Distribution of Traditional Cardiovascular Disease Risk Factors at Follow-up Year 20, According to 3 Indices of Year 20 Body Mass Indexa and Its History, Coronary Artery Risk Development in Young Adults (CARDIA) Study, 1985–2006

 Quintile 1 (n = 448)
 
Quintile 2 (n = 449)
 
Quintile 3 (n = 449)
 
Quintile 4 (n = 449)
 
Quintile 5 (n = 448)
 
P Value 
 Mean (SD) Mean (SD) Mean (SD) Mean (SD) Mean (SD) 
 BMI at Year 0
 
 
 <20.7
 
20.7–<22.5
 
22.5–<24.2
 
24.2–<26.9
 
≥26.9
 
 
Age, years 45.0 (3.6)  44.9 (3.4)  45.1 (3.5)  45.9 (3.5)  45.6 (3.6)  <0.01 
Female sex, %  71.2  58.1  45.9  43.7  59.6 0.02 
White race/ethnicity, %  63.6  66.1  61.7  60.1  41.1 <0.01 
Current smoker, %  16.5  14.9  19.2  19.4  18.3 0.35 
Physical activity score, exercise unitsb 332 (260)  388 (298)  363 (275)  332 (254)  273 (257)  <0.01 
Systolic blood pressure, mm Hg 112.1 (13.9)  113.5 (14.4)  115.2 (13.3)  116.1 (12.7)  119.8 (15.9)  <0.01 
HDL cholesterol, mg/dL 60.5 (18.2)  57.9 (17.0)  53.3 (16.2)  51.2 (15.2)  50.4 (15.4)  <0.01 
LDL cholesterol, mg/dL 108.9 (32.6)  108.9 (29.7)  111.6 (33.0)  113.0 (30.0)  110.3 (30.7)  0.20 
Triglycerides, mg/dL 95.6 (76.0)  96.1 (58.5)  112.9 (83.8)  121.9 (88.8)  123.6 (90.3)  <0.01 
Fasting glucose, mg/dL 92.7 (13.6)  92.8 (11.1)  96.1 (24.7)  98.4 (20.5)  107.1 (36.1)  <0.01 
 Quintile 1 (n = 448)
 
Quintile 2 (n = 449)
 
Quintile 3 (n = 449)
 
Quintile 4 (n = 449)
 
Quintile 5 (n = 448)
 
P Value 
 Mean (SD) Mean (SD) Mean (SD) Mean (SD) Mean (SD) 
 BMI at Year 0
 
 
 <20.7
 
20.7–<22.5
 
22.5–<24.2
 
24.2–<26.9
 
≥26.9
 
 
Age, years 45.0 (3.6)  44.9 (3.4)  45.1 (3.5)  45.9 (3.5)  45.6 (3.6)  <0.01 
Female sex, %  71.2  58.1  45.9  43.7  59.6 0.02 
White race/ethnicity, %  63.6  66.1  61.7  60.1  41.1 <0.01 
Current smoker, %  16.5  14.9  19.2  19.4  18.3 0.35 
Physical activity score, exercise unitsb 332 (260)  388 (298)  363 (275)  332 (254)  273 (257)  <0.01 
Systolic blood pressure, mm Hg 112.1 (13.9)  113.5 (14.4)  115.2 (13.3)  116.1 (12.7)  119.8 (15.9)  <0.01 
HDL cholesterol, mg/dL 60.5 (18.2)  57.9 (17.0)  53.3 (16.2)  51.2 (15.2)  50.4 (15.4)  <0.01 
LDL cholesterol, mg/dL 108.9 (32.6)  108.9 (29.7)  111.6 (33.0)  113.0 (30.0)  110.3 (30.7)  0.20 
Triglycerides, mg/dL 95.6 (76.0)  96.1 (58.5)  112.9 (83.8)  121.9 (88.8)  123.6 (90.3)  <0.01 
Fasting glucose, mg/dL 92.7 (13.6)  92.8 (11.1)  96.1 (24.7)  98.4 (20.5)  107.1 (36.1)  <0.01 
 Slope of BMI Change From Year 0 to Year 20, units/year
 
 
 −1.12 to <0.08
 
0.08–<0.17
 
0.17–<0.25
 
0.25–<0.40
 
0.40–1.16
 
 
Age, years 45.5 (3.2)  46.0 (3.2)  45.5 (3.4)  45.2 (3.7)  44.3 (3.9)  <0.01 
Female sex, %  59.8  46.1  47.2  51.7  73.7 <0.01 
White race/ethnicity, %  66.7  69.9  64.1  51.2  40.6 <0.01 
Current smoker, %  24.3  19.4  16.9  15.1  12.5 <0.01 
Physical activity score, exercise units 394 (276)  374 (264)  333 (277)  300 (250)  285 (278)  <0.01 
Systolic blood pressure, mm Hg 113.1 (15.1)  113.6 (13.8)  115.4 (14.2)  117.8 (13.6)  116.9 (14.2)  <0.01 
HDL cholesterol, mg/dL 63.5 (19.7)  56.9 (16.7)  51.8 (14.9)  50.9 (14.9)  50.2 (13.8)  <0.01 
LDL cholesterol, mg/dL 98.3 (28.8)  111.6 (31.0)  115.3 (31.0)  114.5 (31.5)  112.93 (30.8)  <0.01 
Triglycerides, mg/dL 87.5 (66.3)  100.1 (68.4)  114.5 (82.3)  123.9 (94.3)  124.0 (85.2)  <0.01 
Fasting glucose, mg/dL 96.0 (33.2)  95.6 (21.2)  96.2 (19.1)  99.1 (19.3)  100.1 (21.9)  <0.01 
 Slope of BMI Change From Year 0 to Year 20, units/year
 
 
 −1.12 to <0.08
 
0.08–<0.17
 
0.17–<0.25
 
0.25–<0.40
 
0.40–1.16
 
 
Age, years 45.5 (3.2)  46.0 (3.2)  45.5 (3.4)  45.2 (3.7)  44.3 (3.9)  <0.01 
Female sex, %  59.8  46.1  47.2  51.7  73.7 <0.01 
White race/ethnicity, %  66.7  69.9  64.1  51.2  40.6 <0.01 
Current smoker, %  24.3  19.4  16.9  15.1  12.5 <0.01 
Physical activity score, exercise units 394 (276)  374 (264)  333 (277)  300 (250)  285 (278)  <0.01 
Systolic blood pressure, mm Hg 113.1 (15.1)  113.6 (13.8)  115.4 (14.2)  117.8 (13.6)  116.9 (14.2)  <0.01 
HDL cholesterol, mg/dL 63.5 (19.7)  56.9 (16.7)  51.8 (14.9)  50.9 (14.9)  50.2 (13.8)  <0.01 
LDL cholesterol, mg/dL 98.3 (28.8)  111.6 (31.0)  115.3 (31.0)  114.5 (31.5)  112.93 (30.8)  <0.01 
Triglycerides, mg/dL 87.5 (66.3)  100.1 (68.4)  114.5 (82.3)  123.9 (94.3)  124.0 (85.2)  <0.01 
Fasting glucose, mg/dL 96.0 (33.2)  95.6 (21.2)  96.2 (19.1)  99.1 (19.3)  100.1 (21.9)  <0.01 
 Fluctuation in BMI From Year 0 to Year 20
 
 
 0.11–<0.63
 
0.63–<0.91
 
0.91–<1.21
 
1.21–<1.79
 
1.79–9.07
 
 
Age, years 45.4 (3.3)  45.6 (3.5)  45.3 (3.5)  45.3 (3.5)  44.9 (3.8)  0.03 
Female sex, %  44.9  48.6  53.9  58.1  73.0 <0.01 
White race/ethnicity, %  70.3  63.0  62.4  50.6  46.4 <0.01 
Current smoker, %  14.5  14.5  15.1  21.2  23.0 <0.01 
Physical activity score, exercise units 410 (288)  345 (258)  339 (268)  311 (269)  281 (259)  <0.01 
Systolic blood pressure, mm Hg 112.6 (12.0)  114.6 (14.2)  115.2 (13.9)  118.1 (15.7)  116.3 (14.9)  <0.01 
HDL cholesterol, mg/dL 57.6 (18.4)  56.3 (17.5)  54.3 (16.3)  52.6 (15.2)  52.5 (16.1)  <0.01 
LDL cholesterol, mg/dL 105.0 (30.5)  113.3 (31.8)  111.0 (23.9)  114.1 (33.2)  109.3 (30.9)  <0.01 
Triglycerides, mg/dL 97.7 (74.6)  105.3 (66.1)  111.1 (77.8)  117.4 (97.6)  118.5 (84.8)  <0.01 
Fasting glucose, mg/dL 93.2 (11.5)  95.8 (18.0)  96.3 (19.5)  99.5 (28.7)  102.1 (32.8)  <0.01 
 Fluctuation in BMI From Year 0 to Year 20
 
 
 0.11–<0.63
 
0.63–<0.91
 
0.91–<1.21
 
1.21–<1.79
 
1.79–9.07
 
 
Age, years 45.4 (3.3)  45.6 (3.5)  45.3 (3.5)  45.3 (3.5)  44.9 (3.8)  0.03 
Female sex, %  44.9  48.6  53.9  58.1  73.0 <0.01 
White race/ethnicity, %  70.3  63.0  62.4  50.6  46.4 <0.01 
Current smoker, %  14.5  14.5  15.1  21.2  23.0 <0.01 
Physical activity score, exercise units 410 (288)  345 (258)  339 (268)  311 (269)  281 (259)  <0.01 
Systolic blood pressure, mm Hg 112.6 (12.0)  114.6 (14.2)  115.2 (13.9)  118.1 (15.7)  116.3 (14.9)  <0.01 
HDL cholesterol, mg/dL 57.6 (18.4)  56.3 (17.5)  54.3 (16.3)  52.6 (15.2)  52.5 (16.1)  <0.01 
LDL cholesterol, mg/dL 105.0 (30.5)  113.3 (31.8)  111.0 (23.9)  114.1 (33.2)  109.3 (30.9)  <0.01 
Triglycerides, mg/dL 97.7 (74.6)  105.3 (66.1)  111.1 (77.8)  117.4 (97.6)  118.5 (84.8)  <0.01 
Fasting glucose, mg/dL 93.2 (11.5)  95.8 (18.0)  96.3 (19.5)  99.5 (28.7)  102.1 (32.8)  <0.01 

Abbreviations: BMI, body mass index; HDL, high density lipoprotein; LDL, low density lipoprotein; SD, standard deviation.

a

Weight (kg)/height (m)2.

b

Physical activity score was derived from the CARDIA Physical Activity History questionnaire (17), a simplified version of the Minnesota Leisure Time Physical Activity Questionnaire. Physical activity is expressed in arbitrary exercise units. The theoretical range of the scores is 0 to 2,592; the actual maximum was 1,770.

Baseline BMI had a coefficient for correlation with year 20 BMI of 0.72. Slope of BMI was not associated with baseline BMI. BMI fluctuation was weakly and positively correlated with BMI slope (r = 0.21) but was more strongly associated with BMI at both baseline and year 20 (r = 0.46 for both). The correlation of IMT with Agatston score was 0.10.

Associations between BMI at various time points and CAC or IMT at year 20

Year 0 BMI was positively associated with CAC at year 20 (Table 2) (Ptrend < 0.01). However, the association weakened progressively as the BMI measurement neared the time of CAC measurement, which we illustrate by showing the weaker association of year 10 BMI with CAC. The association became U-shaped at year 15 (data not shown) and deepened further with year 20 BMI (Pquadratic = 0.02). However, BMI at each measurement time was strongly associated with IMT at year 20 (Table 2).

Table 2.

Adjusteda Odds Ratios for Risk of High Coronary Artery Calcification (>90th Percentile) and High Common Carotid Artery Intima-Media Thickness (>90th Percentile) at Follow-up Year 20, According to Category of Body Mass Indexb at Years 0, 10, and 20, Coronary Artery Risk Development in Young Adults (CARDIA) Study, 1985–2006

Outcome and BMI Measurement BMI Category
 
Ptrend Pquadratic 
<22 22–<25 25–<30 30–<35 ≥35 
Coronary artery calcification at year 20        
    BMI at year 0        
        No. of cases/no. of participants 45/762 66/748 83/530 19/144 10/59   
        Unadjusted % 5.9 8.8 15.7 13.2 17.0   
        Model 1 OR (95% CI) Referent 1.1 (0.8, 1.7) 2.3 (1.5, 3.5) 2.3 (1.3, 4.2) 3.7 (1.6, 8.2) <0.01 0.82 
        Model 2 OR (95% CI) Referent 0.9 (0.6, 1.4) 1.7 (1.1, 2.5) 1.3 (0.7, 2.4) 2.1 (0.9, 5.0) 0.02 1.00 
    BMI at year 10        
        No. of cases/no. of participants 40/493 36/536 86/726 42/301 19/187   
        Unadjusted % 8.1 6.7 11.9 14.0 10.2   
        Model 1 OR (95% CI) Referent 0.5 (0.3, 0.8) 1.0 (0.6, 1.4) 1.4 (0.9, 2.3) 1.2 (0.7, 2.3) 0.02 0.16 
        Model 2 OR (95% CI) Referent 0.4 (0.2, 0.7) 0.7 (0.4, 1.1) 0.8 (0.5, 1.4) 0.6 (0.3, 1.3) 0.81 0.18 
    BMI at year 20        
        No. of cases/no. of participants 25/234 33/401 81/804 47/427 37/377   
        Unadjusted % 10.7 8.2 10.1 11.0 9.8   
        Model 1 OR (95% CI) Referent 0.6 (0.3, 1.0) 0.6 (0.4, 1.0) 0.8 (0.5, 1.4) 0.9 (0.5, 1.7) 0.45 0.02 
        Model 2 OR (95% CI) Referent 0.5 (0.3, 1.0) 0.5 (0.3, 0.9) 0.7 (0.4, 1.2) 0.8 (0.4, 1.4) 0.99 0.02 
Intima-media thickness at year 20        
    BMI at year 0        
        No. of cases/no. of participants 29/762 71/748 84/530 31/144 10/59   
        Unadjusted % 3.8 9.5 15.9 21.5 17.0   
        Model 1 OR (95% CI) Referent 2.2 (1.4, 3.4) 3.5 (2.2, 5.5) 5.3 (3.0, 9.5) 3.5 (1.5, 7.9) <0.01 <0.01 
        Model 2 OR (95% CI) Referent 2.0 (1.3, 3.2) 2.8 (1.8, 4.5) 3.8 (2.1, 7.0) 2.3 (0.9, 5.4) <0.01 <0.01 
    BMI at year 10        
        No. of cases/no. of participants 23/493 33/536 84/726 48/301 37/187   
        Unadjusted % 4.7 6.2 11.6 16.0 19.8   
        Model 1 OR (95% CI) Referent 1.1 (0.6, 1.9) 1.8 (1.1, 3.0) 2.8 (1.6, 4.8) 4.1 (2.3, 7.4) <0.01 0.54 
        Model 2 OR (95% CI) Referent 0.9 (0.5, 1.6) 1.3 (0.8, 2.1) 1.6 (0.9, 2.9) 2.1 (1.1, 4.0) <0.01 0.45 
    BMI at year 20        
        No. of cases/no. of participants 6/234 20/401 67/804 65/427 67/377   
        Unadjusted % 2.6 5.0 8.3 15.2 17.8   
        Model 1 OR (95% CI) Referent 1.6 (0.6, 4.2) 2.3 (1.0, 5.5) 4.5 (1.9, 10.9) 7.0 (2.9, 16.8) <0.01 0.80 
        Model 2 OR (95% CI) Referent 1.6 (0.6, 4.3) 1.9 (0.7, 4.8) 3.4 (1.3, 8.8) 4.8 (1.9, 12.4) <0.01 0.71 
Outcome and BMI Measurement BMI Category
 
Ptrend Pquadratic 
<22 22–<25 25–<30 30–<35 ≥35 
Coronary artery calcification at year 20        
    BMI at year 0        
        No. of cases/no. of participants 45/762 66/748 83/530 19/144 10/59   
        Unadjusted % 5.9 8.8 15.7 13.2 17.0   
        Model 1 OR (95% CI) Referent 1.1 (0.8, 1.7) 2.3 (1.5, 3.5) 2.3 (1.3, 4.2) 3.7 (1.6, 8.2) <0.01 0.82 
        Model 2 OR (95% CI) Referent 0.9 (0.6, 1.4) 1.7 (1.1, 2.5) 1.3 (0.7, 2.4) 2.1 (0.9, 5.0) 0.02 1.00 
    BMI at year 10        
        No. of cases/no. of participants 40/493 36/536 86/726 42/301 19/187   
        Unadjusted % 8.1 6.7 11.9 14.0 10.2   
        Model 1 OR (95% CI) Referent 0.5 (0.3, 0.8) 1.0 (0.6, 1.4) 1.4 (0.9, 2.3) 1.2 (0.7, 2.3) 0.02 0.16 
        Model 2 OR (95% CI) Referent 0.4 (0.2, 0.7) 0.7 (0.4, 1.1) 0.8 (0.5, 1.4) 0.6 (0.3, 1.3) 0.81 0.18 
    BMI at year 20        
        No. of cases/no. of participants 25/234 33/401 81/804 47/427 37/377   
        Unadjusted % 10.7 8.2 10.1 11.0 9.8   
        Model 1 OR (95% CI) Referent 0.6 (0.3, 1.0) 0.6 (0.4, 1.0) 0.8 (0.5, 1.4) 0.9 (0.5, 1.7) 0.45 0.02 
        Model 2 OR (95% CI) Referent 0.5 (0.3, 1.0) 0.5 (0.3, 0.9) 0.7 (0.4, 1.2) 0.8 (0.4, 1.4) 0.99 0.02 
Intima-media thickness at year 20        
    BMI at year 0        
        No. of cases/no. of participants 29/762 71/748 84/530 31/144 10/59   
        Unadjusted % 3.8 9.5 15.9 21.5 17.0   
        Model 1 OR (95% CI) Referent 2.2 (1.4, 3.4) 3.5 (2.2, 5.5) 5.3 (3.0, 9.5) 3.5 (1.5, 7.9) <0.01 <0.01 
        Model 2 OR (95% CI) Referent 2.0 (1.3, 3.2) 2.8 (1.8, 4.5) 3.8 (2.1, 7.0) 2.3 (0.9, 5.4) <0.01 <0.01 
    BMI at year 10        
        No. of cases/no. of participants 23/493 33/536 84/726 48/301 37/187   
        Unadjusted % 4.7 6.2 11.6 16.0 19.8   
        Model 1 OR (95% CI) Referent 1.1 (0.6, 1.9) 1.8 (1.1, 3.0) 2.8 (1.6, 4.8) 4.1 (2.3, 7.4) <0.01 0.54 
        Model 2 OR (95% CI) Referent 0.9 (0.5, 1.6) 1.3 (0.8, 2.1) 1.6 (0.9, 2.9) 2.1 (1.1, 4.0) <0.01 0.45 
    BMI at year 20        
        No. of cases/no. of participants 6/234 20/401 67/804 65/427 67/377   
        Unadjusted % 2.6 5.0 8.3 15.2 17.8   
        Model 1 OR (95% CI) Referent 1.6 (0.6, 4.2) 2.3 (1.0, 5.5) 4.5 (1.9, 10.9) 7.0 (2.9, 16.8) <0.01 0.80 
        Model 2 OR (95% CI) Referent 1.6 (0.6, 4.3) 1.9 (0.7, 4.8) 3.4 (1.3, 8.8) 4.8 (1.9, 12.4) <0.01 0.71 

Abbreviations: BMI, body mass index; CI, confidence interval; OR, odds ratio.

a

Model 1 results were adjusted for age, sex, race/ethnicity, study center, cigarette smoking, and physical activity. Model 2 results were further adjusted for systolic blood pressure, high density lipoprotein cholesterol, low density lipoprotein cholesterol, triglycerides, and fasting glucose. Each covariate was measured at the same examination as the corresponding BMI.

b

Weight (kg)/height (m)2.

Associations of 3 indices of 20-year BMI and its history with CAC or IMT at year 20

To enhance understanding of the progressively deepening U-shaped association as the time of BMI measurement approached the time of CAC measurement, we included the 3 indices of year 20 BMI and its history simultaneously in the model. After we considered the effects of slope and fluctuation in BMI during 20 years, BMI at baseline was positively related to year 20 CAC (Table 3) (Ptrend = 0.03); the pattern of association was similar to the one shown in Table 2 without adjustment for changes in BMI. However, the slope of BMI during 20 years (up to 4 units during 20 years) was inversely associated with CAC at year 20, independently of BMI at year 0 and fluctuation in BMI during 20 years (Table 3) (Ptrend < 0.01). Participants in the lowest quintile of linear BMI slope showed the highest risk of CAC at year 20, and those who gained more weight during the 20-year period tended to have a lower risk of CAC (Ptrend < 0.01). BMI fluctuation was significantly and positively associated with CAC at year 20 (Ptrend < 0.01), particularly in the highest quintiles of fluctuation. All of these associations were similarly observed among never smokers (data not shown).

Table 3.

Adjusteda Odds Ratios for Risk of High Coronary Artery Calcification (>90th Percentile) at Follow-up Year 20, According to 3 Indices of Year 20 Body Mass Indexb and Its History, Coronary Artery Risk Development in Young Adults (CARDIA) Study, 1985–2006

 BMI at Year 0
 
Ptrend 
 Q1 (19.8)c Q2 (21.6) Q3 (23.4) Q4 (25.3) Q5 (29.7) 
No. of cases/no. of participants 28/448 25/449 46/449 59/449 65/448  
Unadjusted % 6.3 5.6 10.2 13.1 14.5  
Model 1 OR (95% CI) Referent 0.8 (0.4, 1.4) 1.2 (0.7, 2.0) 1.3 (0.8, 2.2) 1.8 (1.1, 3.1) <0.01 
Model 2 OR (95% CI) Referent 0.8 (0.4, 1.4) 1.1 (0.7, 1.9) 1.3 (0.8, 2.1) 1.6 (0.9, 2.8) 0.03 
 BMI at Year 0
 
Ptrend 
 Q1 (19.8)c Q2 (21.6) Q3 (23.4) Q4 (25.3) Q5 (29.7) 
No. of cases/no. of participants 28/448 25/449 46/449 59/449 65/448  
Unadjusted % 6.3 5.6 10.2 13.1 14.5  
Model 1 OR (95% CI) Referent 0.8 (0.4, 1.4) 1.2 (0.7, 2.0) 1.3 (0.8, 2.2) 1.8 (1.1, 3.1) <0.01 
Model 2 OR (95% CI) Referent 0.8 (0.4, 1.4) 1.1 (0.7, 1.9) 1.3 (0.8, 2.1) 1.6 (0.9, 2.8) 0.03 
 Linear Slope of BMI From Year 0 to Year 20
 
 
 Q1 (0.01) Q2 (0.12) Q3 (0.21) Q4 (0.32) Q5 (0.54)  
No. of cases/no. of participants 66/448 57/449 34/449 40/449 26/448  
Unadjusted % 14.7 12.7 7.6 8.9 5.8  
Model 1 OR (95% CI) Referent 0.7 (0.4, 1.0) 0.4 (0.3, 0.6) 0.5 (0.3, 0.8) 0.4 (0.3, 0.7) <0.01 
Model 2 OR (95% CI) Referent 0.7 (0.4, 1.0) 0.4 (0.2, 0.6) 0.5 (0.3, 0.8) 0.4 (0.2, 0.7) <0.01 
 Linear Slope of BMI From Year 0 to Year 20
 
 
 Q1 (0.01) Q2 (0.12) Q3 (0.21) Q4 (0.32) Q5 (0.54)  
No. of cases/no. of participants 66/448 57/449 34/449 40/449 26/448  
Unadjusted % 14.7 12.7 7.6 8.9 5.8  
Model 1 OR (95% CI) Referent 0.7 (0.4, 1.0) 0.4 (0.3, 0.6) 0.5 (0.3, 0.8) 0.4 (0.3, 0.7) <0.01 
Model 2 OR (95% CI) Referent 0.7 (0.4, 1.0) 0.4 (0.2, 0.6) 0.5 (0.3, 0.8) 0.4 (0.2, 0.7) <0.01 
 Fluctuation in BMI During 20 Years
 
 
 Q1 (0.49) Q2 (0.77) Q3 (1.05) Q4 (1.46) Q5 (2.34)  
No. of cases/no. of participants 36/448 42/449 39/449 61/449 45/448  
Unadjusted % 8.0 9.4 8.7 13.6 10.0  
Model 1 OR (95% CI) Referent 1.2 (0.8, 2.0) 1.2 (0.7, 2.0) 2.1 (1.3, 3.4) 1.7 (1.0, 3.0) <0.01 
Model 2 OR (95% CI) Referent 1.2 (0.7, 1.9) 1.1 (0.7, 1.9) 1.9 (1.2, 3.2) 1.7 (1.0, 3.0) <0.01 
 Fluctuation in BMI During 20 Years
 
 
 Q1 (0.49) Q2 (0.77) Q3 (1.05) Q4 (1.46) Q5 (2.34)  
No. of cases/no. of participants 36/448 42/449 39/449 61/449 45/448  
Unadjusted % 8.0 9.4 8.7 13.6 10.0  
Model 1 OR (95% CI) Referent 1.2 (0.8, 2.0) 1.2 (0.7, 2.0) 2.1 (1.3, 3.4) 1.7 (1.0, 3.0) <0.01 
Model 2 OR (95% CI) Referent 1.2 (0.7, 1.9) 1.1 (0.7, 1.9) 1.9 (1.2, 3.2) 1.7 (1.0, 3.0) <0.01 

Abbreviations: BMI, body mass index; CI, confidence interval; OR, odds ratio; Q, quintile.

a

This table presents results from 2 separate models with different covariates. Each model included BMI at year 0, slope of BMI from year 0 to year 20, fluctuation in BMI from year 0 to year 20, and covariates measured at year 20. Covariates for model 1 were age, sex, race/ethnicity, study center, cigarette smoking, and physical activity. Covariates for model 2 were those for model 1 plus systolic blood pressure, high density lipoprotein cholesterol, low density lipoprotein cholesterol, triglycerides, and fasting glucose.

b

Weight (kg)/height (m)2.

c

Numbers in parentheses, median value.

IMT showed different patterns of association from those of CAC (Table 4). In models including the 3 indices of year 20 BMI and its history, BMI at baseline was strongly and positively related to IMT at year 20 (Ptrend < 0.01). In addition, the slope of BMI during 20 years was positively associated with IMT at year 20 (Ptrend < 0.01). BMI fluctuation during 20 years was not associated with IMT at year 20.

Table 4.

Adjusteda Odds Ratios for Risk of High Common Carotid Artery Intima-Media Thickness (>90th Percentile) at Follow-up Year 20, According to 3 Indices of Body Mass Indexb and Its History, Coronary Artery Risk Development in Young Adults (CARDIA) Study, 1985–2006

 BMI at Year 0
 
Ptrend 
 Q1 (19.8)c Q2 (21.6) Q3 (23.4) Q4 (25.3) Q5 (29.7) 
No. of cases/no. of participants 15/448 21/449 41/449 61/449 87/448  
Unadjusted % 3.3 4.7 9.1 13.6 19.4  
Model 1 OR (95% CI) Referent 1.3 (0.6, 2.5) 2.1 (1.1, 3.9) 2.8 (1.5, 5.1) 4.6 (2.5, 8.4) <0.01 
Model 2 OR (95% CI) Referent 1.2 (0.6, 2.4) 2.0 (1.0, 3.7) 2.6 (1.4, 4.9) 3.6 (2.0, 6.8) <0.01 
 BMI at Year 0
 
Ptrend 
 Q1 (19.8)c Q2 (21.6) Q3 (23.4) Q4 (25.3) Q5 (29.7) 
No. of cases/no. of participants 15/448 21/449 41/449 61/449 87/448  
Unadjusted % 3.3 4.7 9.1 13.6 19.4  
Model 1 OR (95% CI) Referent 1.3 (0.6, 2.5) 2.1 (1.1, 3.9) 2.8 (1.5, 5.1) 4.6 (2.5, 8.4) <0.01 
Model 2 OR (95% CI) Referent 1.2 (0.6, 2.4) 2.0 (1.0, 3.7) 2.6 (1.4, 4.9) 3.6 (2.0, 6.8) <0.01 
 Linear Slope of BMI From Year 0 to Year 20
 
 
 Q1 (0.01) Q2 (0.12) Q3 (0.21) Q4 (0.32) Q5 (0.54)  
No. of cases/no. of participants 35/448 27/449 48/449 58/449 57/448  
Unadjusted % 7.8 6.0 10.7 12.9 12.7  
Model 1 OR (95% CI) Referent 0.7 (0.4, 1.2) 1.4 (0.9, 2.4) 1.7 (1.0, 2.7) 1.9 (1.1, 3.1) <0.01 
Model 2 OR (95% CI) Referent 0.7 (0.4, 1.2) 1.4 (0.8, 2.3) 1.6 (0.9, 2.6) 1.8 (1.0, 3.0) <0.01 
 Linear Slope of BMI From Year 0 to Year 20
 
 
 Q1 (0.01) Q2 (0.12) Q3 (0.21) Q4 (0.32) Q5 (0.54)  
No. of cases/no. of participants 35/448 27/449 48/449 58/449 57/448  
Unadjusted % 7.8 6.0 10.7 12.9 12.7  
Model 1 OR (95% CI) Referent 0.7 (0.4, 1.2) 1.4 (0.9, 2.4) 1.7 (1.0, 2.7) 1.9 (1.1, 3.1) <0.01 
Model 2 OR (95% CI) Referent 0.7 (0.4, 1.2) 1.4 (0.8, 2.3) 1.6 (0.9, 2.6) 1.8 (1.0, 3.0) <0.01 
 Fluctuation in BMI During 20 Years
 
 
 Q1 (0.49) Q2 (0.77) Q3 (1.05) Q4 (1.46) Q5 (2.34)  
No. of cases/no. of participants 31/448 41/449 42/449 50/449 61/448  
Unadjusted % 6.9 9.1 9.4 11.1 13.6  
Model 1 OR (95% CI) Referent 1.0 (0.6, 1.7) 1.0 (0.6, 1.7) 1.0 (0.6, 1.7) 1.3 (0.8, 2.2) 0.30 
Model 2 OR (95% CI) Referent 0.9 (0.5, 1.6) 1.0 (0.6, 1.6) 0.9 (0.5, 1.5) 1.2 (0.7, 2.1) 0.55 
 Fluctuation in BMI During 20 Years
 
 
 Q1 (0.49) Q2 (0.77) Q3 (1.05) Q4 (1.46) Q5 (2.34)  
No. of cases/no. of participants 31/448 41/449 42/449 50/449 61/448  
Unadjusted % 6.9 9.1 9.4 11.1 13.6  
Model 1 OR (95% CI) Referent 1.0 (0.6, 1.7) 1.0 (0.6, 1.7) 1.0 (0.6, 1.7) 1.3 (0.8, 2.2) 0.30 
Model 2 OR (95% CI) Referent 0.9 (0.5, 1.6) 1.0 (0.6, 1.6) 0.9 (0.5, 1.5) 1.2 (0.7, 2.1) 0.55 

Abbreviations: BMI, body mass index; CI, confidence interval; OR, odds ratio; Q, quintile.

a

This table presents results from 2 separate models with different covariates. Each model included BMI at year 0, slope of BMI from year 0 to year 20, fluctuation in BMI from year 0 to year 20, and covariates measured at year 20. Covariates for model 1 were age, sex, race/ethnicity, study center, cigarette smoking, and physical activity. Covariates for model 2 were those for model 1 plus systolic blood pressure, high density lipoprotein cholesterol, low density lipoprotein cholesterol, triglycerides, and fasting glucose.

b

Weight (kg)/height (m)2.

c

Numbers in parentheses, median value.

Associations between 3 indices of year 15 BMI and its history and progression of CAC

Table 5 presents associations of the progression of CAC from year 15 to year 20 with the 3 indices of year 15 BMI. In these progression analyses, the trends were similar to those in Table 3. The inverse association of the slope of BMI during 15 years and the positive association of the fluctuation in BMI during 15 years were strengthened when we used stricter definitions of CAC progression (Web Table 4).

Table 5.

Adjusteda Odds Ratios for Progression of Coronary Artery Calcification From Follow-up Year 15 to Year 20, According to 3 Indices of Year 15 Body Mass Indexb and Its History, Coronary Artery Risk Development in Young Adults (CARDIA) Study, 1985–2006

 BMI at Year 0
 
Ptrend 
 Q1 (19.8)c Q2 (21.6) Q3 (23.4) Q4 (25.3) Q5 (29.7) 
No. of cases/no. of participants 24/448 25/449 40/449 56/449 62/448  
Unadjusted % 5.4 5.6 8.9 12.5 13.8  
Model 1 OR (95% CI) Referent 0.9 (0.5, 1.6) 1.1 (0.6, 1.9) 1.4 (0.8, 2.4) 2.0 (1.1, 3.4) <0.01 
Model 2 OR (95% CI) Referent 0.8 (0.4, 1.5) 0.9 (0.5, 1.7) 1.2 (0.7, 2.0) 1.5 (0.8-2.6) 0.08 
 BMI at Year 0
 
Ptrend 
 Q1 (19.8)c Q2 (21.6) Q3 (23.4) Q4 (25.3) Q5 (29.7) 
No. of cases/no. of participants 24/448 25/449 40/449 56/449 62/448  
Unadjusted % 5.4 5.6 8.9 12.5 13.8  
Model 1 OR (95% CI) Referent 0.9 (0.5, 1.6) 1.1 (0.6, 1.9) 1.4 (0.8, 2.4) 2.0 (1.1, 3.4) <0.01 
Model 2 OR (95% CI) Referent 0.8 (0.4, 1.5) 0.9 (0.5, 1.7) 1.2 (0.7, 2.0) 1.5 (0.8-2.6) 0.08 
 Linear Slope of BMI From Year 0 to Year 15
 
 
 Q1 (−0.01) Q2 (0.12) Q3 (0.22) Q4 (0.36) Q5 (0.61) 
No. of cases/no. of participants 51/448 50/449 43/449 41/449 22/448  
Unadjusted % 11.4 11.1 9.6 9.1 4.9  
Model 1 OR (95% CI) Referent 1.0 (0.6, 1.5) 0.7 (0.4, 1.1) 0.7 (0.4, 1.1) 0.4 (0.3, 0.8) <0.01 
Model 2 OR (95% CI) Referent 0.9 (0.6, 1.4) 0.6 (0.4, 1.0) 0.6 (0.3, 0.9) 0.4 (0.2, 0.6) <0.01 
 Linear Slope of BMI From Year 0 to Year 15
 
 
 Q1 (−0.01) Q2 (0.12) Q3 (0.22) Q4 (0.36) Q5 (0.61) 
No. of cases/no. of participants 51/448 50/449 43/449 41/449 22/448  
Unadjusted % 11.4 11.1 9.6 9.1 4.9  
Model 1 OR (95% CI) Referent 1.0 (0.6, 1.5) 0.7 (0.4, 1.1) 0.7 (0.4, 1.1) 0.4 (0.3, 0.8) <0.01 
Model 2 OR (95% CI) Referent 0.9 (0.6, 1.4) 0.6 (0.4, 1.0) 0.6 (0.3, 0.9) 0.4 (0.2, 0.6) <0.01 
 Fluctuation in BMI During 15 Years
 
 
 Q1 (0.40) Q2 (0.69) Q3 (0.96) Q4 (1.35) Q5 (2.17)  
No. of cases/no. of participants 31/448 35/449 42/449 51/449 48/448  
Unadjusted % 6.9 7.8 9.4 11.4 10.7  
Model 1 OR (95% CI) Referent 1.1 (0.7, 1.9) 1.5 (0.9, 2.5) 2.1 (1.2, 3.5) 2.2 (1.3, 3.9) <0.01 
Model 2 OR (95% CI) Referent 1.0 (0.6, 1.8) 1.4 (0.8, 2.3) 1.9 (1.1, 3.3) 2.0 (1.1, 3.6) <0.01 
 Fluctuation in BMI During 15 Years
 
 
 Q1 (0.40) Q2 (0.69) Q3 (0.96) Q4 (1.35) Q5 (2.17)  
No. of cases/no. of participants 31/448 35/449 42/449 51/449 48/448  
Unadjusted % 6.9 7.8 9.4 11.4 10.7  
Model 1 OR (95% CI) Referent 1.1 (0.7, 1.9) 1.5 (0.9, 2.5) 2.1 (1.2, 3.5) 2.2 (1.3, 3.9) <0.01 
Model 2 OR (95% CI) Referent 1.0 (0.6, 1.8) 1.4 (0.8, 2.3) 1.9 (1.1, 3.3) 2.0 (1.1, 3.6) <0.01 

Abbreviations: BMI, body mass index; CI, confidence interval; OR, odds ratio; Q, quintile.

a

This table presents results from 2 separate models with different covariates. Each model included BMI at year 0, slope of BMI from year 0 to year 15, fluctuation in BMI from year 0 to year 15, and covariates measured at year 15. Covariates for model 1 were age, sex, race/ethnicity, study center, cigarette smoking, and physical activity. Covariates for model 2 were those for model 1 plus systolic blood pressure, high density lipoprotein cholesterol, low density lipoprotein cholesterol, triglycerides, and fasting glucose.

b

Weight (kg)/height (m)2.

c

Numbers in parentheses, median value.

Sensitivity analyses

Within the lowest quintile of linear slope of BMI, prevalence and progression of CAC were greater for participants who lost weight than for those who gained a small amount of weight (data not shown). Analyses conducted only among subjects with weight gain (weight losers excluded) also showed similar inverse associations (data not shown). Additional adjustment for use of statins and bariatric surgery had little effect on adjusted estimates (data not shown). Generally, results based on waist circumference were similar to those based on BMI (Web Tables 58).

Approximately 40% of weight loss was stated to be intentional when participants were asked at years 7 and 10. Losing weight between years 5 and 7 was associated with a higher risk of CAC at year 20 irrespective of intentionality, but persons who lost weight intentionally between years 7 and 10 had an even higher risk of CAC at year 20 (23.1%) than those who lost weight unintentionally (15.5%).

DISCUSSION

The positive association of year 0 BMI with year 20 CAC progressively deepened in a U shape as the time of BMI measurement approached that of CAC, while IMT was consistently and positively associated with BMI irrespective of the time of BMI measurement. To enhance our understanding of this finding, we used identical analytic approaches to examine comprehensive associations of weight dynamics with CAC and IMT. Weight change and weight fluctuation showed different results depending on which subclinical marker of atherosclerosis was used as the outcome, including the unexpected findings that weight loss, modest weight gain, and greater weight fluctuation were associated with increased CAC at year 20 and CAC progression between years 15 and 20. Because BMI at year 20 can be thought of as the combination of BMI at year 0 with change in BMI over 20 years, this progression towards a nonlinear relation suggests the presence of countervailing pathways linking baseline BMI to higher prevalence and progression of CAC but increase in BMI to lower prevalence and progression of CAC. Although the positive association with baseline BMI may reflect the harmfulness of obesity itself, the inverse association with changes in BMI may reflect other aspects which are closely related to changes in amounts of adipose tissue, as we discuss below. In contrast to CAC, linear change in BMI during 20 years showed positive associations with IMT. The association between baseline BMI and IMT was also stronger than that of CAC, while fluctuation in BMI was not associated with IMT.

The inverse association between linear weight change and CAC may be surprising, but it is not unprecedented. In several observational studies, investigators have reported that weight loss was associated with increased risk of cardiovascular disease or total mortality, while mild-to-moderate weight gain did not increase or even decreased the risk of coronary events (19–21). The increased risk among weight losers has often been interpreted as a reflection of reverse causality due to unintentional weight loss from preexisting illness or cigarette smoking, or loss of lean mass rather than fat mass at older ages (22–24). Decreased risk with some weight gain has been dismissed as biologically implausible (23, 24). However, it is difficult to attribute our CAC findings on weight change to bias due to reverse causality, considering this generally healthy cohort who were young at baseline and successfully aged for 20 years (cumulative death rate = 4%; clinical heart disease was rare). First of all, the inverse trend was clearly observed across a broad range of weight changes, not just for the small group of participants who lost weight. In addition, all of these associations were similarly observed among never smokers.

It is also provocative that weight fluctuation appeared to increase CAC risk. It is generally thought that intentional weight loss is good for health, while unintentional weight loss is harmful. However, difficulties in maintaining reduced body weight after intentional weight loss commonly lead to repeated attempts and weight fluctuation. Thus, our findings on fluctuation suggest that even intentional weight loss may not always be good, if a person does not successfully maintain the intentional weight loss. An adverse effect of weight fluctuation has been regarded as biologically plausible, because weight fluctuation and associated adipose tissue remodeling can have detrimental effects on various cardiovascular disease risk factors (25). Similarly to our studies, weight fluctuation has been linked to some harmful consequences, including cardiovascular disease, in other studies (15, 26).

Many researchers think that higher risk of cardiovascular disease among weight losers and lower risk among weight gainers is not biologically plausible (23, 24). However, there are pathways that may be relevant; for example, lipophilic xenobiotics stored in adipose tissue may explain this surprising finding. In fact, in recent epidemiologic studies, investigators have reported that these xenobiotics were associated with various diseases, including cardiovascular disease in the US general population (27). These xenobiotics can be released from adipose tissue during weight loss, be redistributed to some critical organs, and cause some detrimental effects in vivo (28). On the other hand, weight gain could decrease disease risk by diluting or sequestering the xenobiotics, which are then less available for distribution to critical organs. In fact, in experimental studies, investigators have observed that polychlorinated biphenyls, one typical example of lipophilic xenobiotics, can disturb calcium homeostasis (29, 30). Thus, CAC may be more specifically associated with exposure to lipophilic xenobiotics than are other manifestations of subclinical cardiovascular disease. However, there may be other, unknown biologic mechanisms.

This study had several limitations. First, we could not assess whether the associations observed would hold for well-documented clinical outcomes such as incidence of coronary heart disease or stroke, for which we would need longer follow-up. Second, these analyses addressed only calcified coronary plaque; while carotid plaque may be noncalcified, other, possibly noncalcified coronary arterial changes should be studied separately. Third, these findings may not translate to older persons. Finally, as in all observational studies, residual unidentified confounding is possible.

In conclusion, this prospective study with repeated measures of body weight showed that associations with weight dynamics varied depending on which marker was used, with thicker IMT being associated with weight gain but greater coronary calcification being associated with weight loss, modest weight gain, and weight fluctuation.

Abbreviations

    Abbreviations
  • BMI

    body mass index

  • CAC

    coronary artery calcification

  • CARDIA

    Coronary Artery Risk Development in Young Adults

  • IMT

    intima-media thickness

  • SD

    standard deviation

Author affiliations: Department of Preventive Medicine, School of Medicine, Kyungpook National University, Daegu, South Korea (Duk-Hee Lee); Division of Epidemiology and Community Health, School of Public Health, University of Minnesota, Minneapolis, Minnesota (Duk-Hee Lee, Kyong Park, David R. Jacobs, Jr.); Department of Laboratory Medicine and Pathology, School of Medicine, University of Minnesota, Minneapolis, Minnesota (Michael W. Steffes, Myron Gross); Department of Food and Nutrition, College of Home Economics, Yeungnam University, Daegu, South Korea (Kyong Park); Atherosclerosis and Metabolism Unit, Department of Cardiovascular Diseases, Faculty of Medicine, Katholieke Universiteit, Leuven, Belgium (Paul Holvoet); Department of Quantitative Health Sciences, University of Massachusetts Medical School, Worcester, Massachusetts (Catarina I. Kiefe); Division of Preventive Medicine, School of Medicine, University of Alabama at Birmingham, Birmingham, Alabama (Cora E. Lewis); and Department of Nutrition, Faculty of Medicine, University of Oslo, Oslo, Norway (David R. Jacobs, Jr.).

This work was supported in part by the following National Heart, Lung, and Blood Institute contracts for the CARDIA Study: N01-HC-48047, N01-HC-48048, N01-HC-48049, and N01-HC-48050 (clinics), N01-HC-95095 (coordinating center), and N01-HC-05187 and HHSN268200425205C (computed tomography reading centers).

Conflict of interest: none declared.

References

1.
Kotsis
VT
Stabouli
SV
Papamichael
CM
, et al.  . 
Impact of obesity in intima media thickness of carotid arteries
Obesity (Silver Spring)
 , 
2006
, vol. 
14
 
10
(pg. 
1708
-
1715
)
2.
Lo
J
Dolan
SE
Kanter
JR
, et al.  . 
Effects of obesity, body composition, and adiponectin on carotid intima-media thickness in healthy women
J Clin Endocrinol Metab
 , 
2006
, vol. 
91
 
5
(pg. 
1677
-
1682
)
3.
Oren
A
Vos
LE
Uiterwaal
CS
, et al.  . 
Change in body mass index from adolescence to young adulthood and increased carotid intima-media thickness at 28 years of age: the Atherosclerosis Risk in Young Adults Study
Int J Obes Relat Metab Disord
 , 
2003
, vol. 
27
 
11
(pg. 
1383
-
1390
)
4.
Raitakari
OT
Juonala
M
Kähönen
M
, et al.  . 
Cardiovascular risk factors in childhood and carotid artery intima-media thickness in adulthood: the Cardiovascular Risk in Young Finns Study
JAMA
 , 
2003
, vol. 
290
 
17
(pg. 
2277
-
2283
)
5.
Cassidy
AE
Bielak
LF
Zhou
Y
, et al.  . 
Progression of subclinical coronary atherosclerosis: does obesity make a difference?
Circulation
 , 
2005
, vol. 
111
 
15
(pg. 
1877
-
1882
)
6.
Fox
CS
Hwang
SJ
Massaro
JM
, et al.  . 
Relation of subcutaneous and visceral adipose tissue to coronary and abdominal aortic calcium (from the Framingham Heart Study)
Am J Cardiol
 , 
2009
, vol. 
104
 
4
(pg. 
543
-
547
)
7.
Lee
CD
Jacobs
DR
Jr.
Schreiner
PJ
, et al.  . 
Abdominal obesity and coronary artery calcification in young adults: the Coronary Artery Risk Development in Young Adults (CARDIA) Study
Am J Clin Nutr
 , 
2007
, vol. 
86
 
1
(pg. 
48
-
54
)
8.
Loria
CM
Liu
K
Lewis
CE
, et al.  . 
Early adult risk factor levels and subsequent coronary artery calcification: the CARDIA Study
J Am Coll Cardiol
 , 
2007
, vol. 
49
 
20
(pg. 
2013
-
2020
)
9.
Mahoney
LT
Burns
TL
Stanford
W
, et al.  . 
Coronary risk factors measured in childhood and young adult life are associated with coronary artery calcification in young adults: the Muscatine Study
J Am Coll Cardiol
 , 
1996
, vol. 
27
 
2
(pg. 
277
-
284
)
10.
Mazzone
T
Meyer
PM
Kondos
GT
, et al.  . 
Relationship of traditional and nontraditional cardiovascular risk factors to coronary artery calcium in type 2 diabetes
Diabetes
 , 
2007
, vol. 
56
 
3
(pg. 
849
-
855
)
11.
See
R
Abdullah
SM
McGuire
DK
, et al.  . 
The association of differing measures of overweight and obesity with prevalent atherosclerosis: the Dallas Heart Study
J Am Coll Cardiol
 , 
2007
, vol. 
50
 
8
(pg. 
752
-
759
)
12.
Snell-Bergeon
JK
Hokanson
JE
Kinney
GL
, et al.  . 
Measurement of abdominal fat by CT compared to waist circumference and BMI in explaining the presence of coronary calcium
Int J Obes Relat Metab Disord
 , 
2004
, vol. 
28
 
12
(pg. 
1594
-
1599
)
13.
Dyer
AR
Stamler
J
Greenland
P
Associations of weight change and weight variability with cardiovascular and all-cause mortality in the Chicago Western Electric Company Study
Am J Epidemiol
 , 
2000
, vol. 
152
 
4
(pg. 
324
-
333
)
14.
Park
K
Lee
DH
Erickson
DJ
, et al.  . 
Association of long-term change in waist circumference with insulin resistance
Obesity (Silver Spring)
 , 
2010
, vol. 
18
 
2
(pg. 
370
-
376
)
15.
Iribarren
C
Sharp
DS
Burchfiel
CM
, et al.  . 
Association of weight loss and weight fluctuation with mortality among Japanese American men
N Engl J Med
 , 
1995
, vol. 
333
 
11
(pg. 
686
-
692
)
16.
Friedman
GD
Cutter
GR
Donahue
RP
, et al.  . 
CARDIA: study design, recruitment, and some characteristics of the examined subjects
J Clin Epidemiol
 , 
1988
, vol. 
41
 
11
(pg. 
1105
-
1116
)
17.
Jacobs
DR
Jr
Hahn
L
Haskell
WL
, et al.  . 
Validity and reliability of a short physical activity history: CARDIA Study and Minnesota Heart Health Program
J Cardiopulm Rehabil
 , 
1989
, vol. 
9
 (pg. 
448
-
459
)
18.
Carr
JJ
Nelson
JC
Wong
ND
, et al.  . 
Calcified coronary artery plaque measurement with cardiac CT in population-based studies: standardized protocol of Multi-Ethnic Study of Atherosclerosis (MESA) and Coronary Artery Risk Development in Young Adults (CARDIA) Study
Radiology
 , 
2005
, vol. 
234
 
1
(pg. 
35
-
43
)
19.
Andres
R
Muller
DC
Sorkin
JD
Long-term effects of change in body weight on all-cause mortality. A review
Ann Intern Med
 , 
1993
, vol. 
119
 
7
(pg. 
737
-
743
)
20.
Maru
S
van der Schouw
YT
Gimbrère
CH
, et al.  . 
Body mass index and short-term weight change in relation to mortality in Dutch women after age 50 y
Am J Clin Nutr
 , 
2004
, vol. 
80
 
1
(pg. 
231
-
236
)
21.
Yarnell
JW
Patterson
CC
Thomas
HF
, et al.  . 
Comparison of weight in middle age, weight at 18 years, and weight change between, in predicting subsequent 14 year mortality and coronary events: Caerphilly Prospective Study
J Epidemiol Community Health
 , 
2000
, vol. 
54
 
5
(pg. 
344
-
348
)
22.
Allison
DB
Zannolli
R
Faith
MS
, et al.  . 
Weight loss increases and fat loss decreases all-cause mortality rate: results from two independent cohort studies
Int J Obes Relat Metab Disord
 , 
1999
, vol. 
23
 
6
(pg. 
603
-
611
)
23.
Astrup
A
Weight loss and increased mortality: epidemiologists blinded by observations?
Obes Rev.
 , 
2003
, vol. 
4
 
1
(pg. 
1
-
2
)
24.
Stampfer
M
Weight loss and mortality: what does the evidence show?
PLoS Med
 , 
2005
, vol. 
2
 
6
 
e181. (doi: 10.1371/journal.pmed.0020181)
25.
Montani
JP
Viecelli
AK
Prevot
A
, et al.  . 
Weight cycling during growth and beyond as a risk factor for later cardiovascular diseases: the ‘repeated overshoot’ theory
Int J Obes
 , 
2006
, vol. 
30
 
suppl 4
(pg. 
S58
-
S66
)
26.
Hamm
P
Shekelle
RB
Stamler
J
Large fluctuations in body weight during young adulthood and twenty-five-year risk of coronary death in men
Am J Epidemiol
 , 
1989
, vol. 
129
 
2
(pg. 
312
-
318
)
27.
Ha
MH
Lee
DH
Jacobs
DR
Jr
Association between serum concentrations of persistent organic pollutants and self-reported cardiovascular disease prevalence: results from the National Health and Nutrition Examination Survey, 1999–2002
Environ Health Perspect
 , 
2007
, vol. 
115
 
8
(pg. 
1204
-
1209
)
28.
Imbeault
P
Chevrier
J
Dewailly
E
, et al.  . 
Increase in plasma pollutant levels in response to weight loss in humans is related to in vitro subcutaneous adipocyte basal lipolysis
Int J Obes Relat Metab Disord
 , 
2001
, vol. 
25
 
11
(pg. 
1585
-
1591
)
29.
Fischer
LJ
Wagner
MA
Madhukar
BV
Potential involvement of calcium, CaM kinase II, and MAP kinases in PCB-stimulated insulin release from RINm5F cells
Toxicol Appl Pharmacol
 , 
1999
, vol. 
159
 
3
(pg. 
194
-
203
)
30.
Kang
JH
Park
IS
Oh
WY
, et al.  . 
Inhibition of aroclor 1254-induced depletion of stored calcium prevents the cell death in catecholaminergic cells
Toxicology
 , 
2004
, vol. 
200
 
2-3
(pg. 
93
-
101
)