Methods
Participants were drawn from the Healthy Aging in Neighborhoods of Diversity across the Life Span (HANDLS), a longitudinal population-based study of health disparities based on an area probability sample of Baltimore City initiated in 2004. HANDLS was designed to disentangle the effects of race and SES on risk factors for morbidity and mortality, to examine the incidence and progression of pre-clinical disease, and to follow-up the development and persistence of health disparities, longitudinal health status, and health risks.{Evans, 2004} Participants were examined initially in 2004-2009 and first follow-up examinations began in 2009, approximately 4-5 years after the initial examination.
In the present study we examined 1559 African Americans and whites with mean age of 48.3 years of age at initial examination who have thus far completed two DXA assessments.
Densitometry measurements. Describe DXA.
The first examination measured hip and lumber BMD by DXA (Lunar DPX-iq). The second examination measured BMD by DXA (Hologic QDR Discovery-A). Bland-Altman statistics for cross-calibration between the Lunar and Hologic machines showed a strong correspondence between the two devices.
Bone density measurements were obtained for the total body, hip and lumbar spine.
131 participants with osteophytes were excluded from these analyses.
Covariates.
Smoking and daily alcohol consumption were obtained through self-report.
Race/ethnicity was determined by self-identification. Two racial/ethnic groups were included in this study African American and white.
Analyses. We log-transformed hip and lumber BMD to normalize their distributions and to facilitate interpretation of percent change over time. We performed separate analyses of change in hip and lumber BMD using mixed-model regressions because this technique accounts appropriately for intra-individual correlations over time.{Singer, 2003} We evaluated the effects of race (African-American and white) and socio-economic status (below and above poverty status) on BMD, and adjusted for the effects of age, sex, alcohol consumption, present cigarette smoking, and body mass index. Coefficients from analyses of outcomes transformed by the natural log are interpretable as the percent change in average value of the outcome per unit change in the predictor according to the formula 100(exp(b)-1), where b is a fixed-effect coefficient.{Singer, 2003; Vittinghoff, 2005} We performed analyses with R 3.0.0{R Development Core Team, 2013} and we considered p<.05 for significant differences.
Results
In this subsample of HANDLS, participants who had family incomes above the Federal poverty limit were a year older (p<.05) than participants below the poverty limit (Table 1). There was significantly greater body mass index and fewer current cigarette smokers among those who were above the Federal poverty limit compared with those below the poverty limit. There were no age differences or differences in sex distributions between African Americans and whites, but there were significantly greater proportion of African Americans who were below the poverty limit and were current cigarette smokers. There were no differences in the distribution of sex by poverty status or race. There were also no differences in body mass index associated with race, and neither poverty status nor race was associated with current alcohol.
Table 1. Sample characteristics at initial assessment separately by poverty status and race.
N Mean SD Min Max
Age 1559 48.33 8.94 30.02 66.22 Age at visit
DXAlumbTBMD 1559 1.15 0.19 0.60 2.15 DXA Lumbar vertebra total bone mineral density [g/cm^2]
DXAhipTotalBMD 1559 1.09 0.17 0.57 2.13 DXA Hip total bone mineral density [g/cm^2]
PhysBMI 1556 29.56 7.05 14.36 56.08 PhysExam-BodyComp: Body mass index
PovStat Age.n Age.mean Age.sd Age.min Age.max
1 Above 952 48.74 9.022 30.02 66.22
2 Below 607 47.69 8.780 30.10 64.99
Welch Two Sample t-test
data: Age by PovStat
t = 2.291, df = 1316, p-value = 0.02215
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
0.1516 1.9603
sample estimates:
mean in group Above mean in group Below
48.74 47.69
Sex
PovStat Women Men
Above "0.59" "0.41"
Below "0.61" "0.39"
Call: xtabs(formula = ~PovStat + Sex, data = dxa[dxa$HNDwave == 1,
])
Number of cases in table: 1559
Number of factors: 2
Test for independence of all factors:
Chisq = 0.6, df = 1, p-value = 0.4
Race
PovStat White AfrAm
Above "0.49" "0.51"
Below "0.31" "0.69"
Call: xtabs(formula = ~PovStat + Race, data = dxa[dxa$HNDwave == 1,
])
Number of cases in table: 1559
Number of factors: 2
Test for independence of all factors:
Chisq = 46, df = 1, p-value = 0.00000000001
PovStat PhysBMI.n PhysBMI.mean PhysBMI.sd PhysBMI.min PhysBMI.max
1 Above 951 29.93 6.929 14.36 56.08
2 Below 605 28.99 7.203 14.70 54.65
Welch Two Sample t-test
data: PhysBMI by PovStat
t = 2.544, df = 1249, p-value = 0.01108
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
0.2149 1.6631
sample estimates:
mean in group Above mean in group Below
29.93 28.99
MedHxAlcCurr
PovStat No Yes
Above "0.76" "0.24"
Below "0.75" "0.25"
Call: xtabs(formula = ~PovStat + MedHxAlcCurr, data = dxa[dxa$HNDwave ==
1, ])
Number of cases in table: 1437
Number of factors: 2
Test for independence of all factors:
Chisq = 0.05, df = 1, p-value = 0.8
MedHxCigaretteCurr
PovStat No Yes
Above "0.61" "0.39"
Below "0.40" "0.60"
Call: xtabs(formula = ~PovStat + MedHxCigaretteCurr, data = dxa[dxa$HNDwave ==
1, ])
Number of cases in table: 1482
Number of factors: 2
Test for independence of all factors:
Chisq = 61, df = 1, p-value = 6e-15
Race Age.n Age.mean Age.sd Age.min Age.max
1 White 654 48.44 8.964 30.10 66.22
2 AfrAm 905 48.25 8.928 30.02 65.98
Welch Two Sample t-test
data: Age by Race
t = 0.4201, df = 1404, p-value = 0.6745
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-0.708 1.094
sample estimates:
mean in group White mean in group AfrAm
48.44 48.25
Sex
Race Women Men
White "0.60" "0.40"
AfrAm "0.59" "0.41"
Call: xtabs(formula = ~Race + Sex, data = dxa[dxa$HNDwave == 1, ])
Number of cases in table: 1559
Number of factors: 2
Test for independence of all factors:
Chisq = 0.12, df = 1, p-value = 0.7
PovStat
Race Above Below
White "0.71" "0.29"
AfrAm "0.54" "0.46"
Call: xtabs(formula = ~Race + PovStat, data = dxa[dxa$HNDwave == 1,
])
Number of cases in table: 1559
Number of factors: 2
Test for independence of all factors:
Chisq = 46, df = 1, p-value = 0.00000000001
Race PhysBMI.n PhysBMI.mean PhysBMI.sd PhysBMI.min PhysBMI.max
1 White 653 29.43 7.006 14.36 56.08
2 AfrAm 903 29.66 7.083 15.21 55.36
Welch Two Sample t-test
data: PhysBMI by Race
t = -0.6354, df = 1414, p-value = 0.5253
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-0.9389 0.4795
sample estimates:
mean in group White mean in group AfrAm
29.43 29.66
MedHxAlcCurr
Race No Yes
White "0.75" "0.25"
AfrAm "0.76" "0.24"
Call: xtabs(formula = ~Race + MedHxAlcCurr, data = dxa[dxa$HNDwave ==
1, ])
Number of cases in table: 1437
Number of factors: 2
Test for independence of all factors:
Chisq = 0.06, df = 1, p-value = 0.8
MedHxCigaretteCurr
Race No Yes
White "0.58" "0.42"
AfrAm "0.50" "0.50"
Call: xtabs(formula = ~Race + MedHxCigaretteCurr, data = dxa[dxa$HNDwave ==
1, ])
Number of cases in table: 1482
Number of factors: 2
Test for independence of all factors:
Chisq = 8, df = 1, p-value = 0.004
In repeated measures analyses age, sex, race, body mass index, and current alcohol use were associated with bone mineral density in the hip (Table 2), and there were significant interactions between age by sex and age by race. Hip BMD was not associated with poverty status or current cigarette smoking. Overall, there was a significant decline in hip BMD over time and whites and African Americans declined at significantly different rates as indicated by a significant interaction between age and race. There were no differences in rates of decline by poverty status. Whites declined by 4.7% per decade; African Americans declined by 2.8% per decade (Figure 1).
Age, sex, race, poverty status, and body mass index were associated with bone mineral density in the lumbar spine (Table 2), and there was a significant interaction between age and sex. Lumbar BMD was not associated with current alcohol consumption or current smoking. Overall, there was a significant decline in lumbar BMD over time and men and women declined at significantly different rates as indicated by a significant interaction between age and sex. There were no differences in the rates of decline in lumbar spine poverty status. There was no difference in the rate of decline by race: Whites declined by 7.3% per decade; African Americans declined by 7.1% per decade (Figure 2).
Table 2. Association of log transformed hip and lumbar bone mineral density with age, sex, race, poverty status, body mass index, current alcohol use, and current cigarette smoking.
Hip
Simultaneous Tests for General Linear Hypotheses
Fit: lmer(formula = logHip ~ ageDecade * (Sex + Race) + PovStat +
PhysBMI + MedHxAlcCurr + MedHxCigaretteCurr + (ageDecade |
HNDid), data = dxa, na.action = na.omit)
Linear Hypotheses:
Estimate Std. Error z value Pr(>|z|)
(Intercept) == 0 -0.277360 0.016423 -16.89 < 2e-16
ageDecade == 0 -0.047679 0.005001 -9.53 < 2e-16
SexMen == 0 0.064858 0.006536 9.92 < 2e-16
RaceAfrAm == 0 0.054689 0.006459 8.47 < 2e-16
PovStatBelow == 0 -0.003313 0.006646 -0.50 0.61819
PhysBMI == 0 0.010132 0.000475 21.31 < 2e-16
MedHxAlcCurrYes == 0 -0.018533 0.007341 -2.52 0.01158
MedHxCigaretteCurrYes == 0 -0.010009 0.006609 -1.51 0.12993
ageDecade:SexMen == 0 0.013651 0.005791 2.36 0.01840
ageDecade:RaceAfrAm == 0 0.019259 0.005789 3.33 0.00088
(Univariate p values reported)
Spine
Simultaneous Tests for General Linear Hypotheses
Fit: lmer(formula = logLum ~ ageDecade * (Sex + Race) + PovStat +
PhysBMI + MedHxAlcCurr + MedHxCigaretteCurr + (ageDecade |
HNDid), data = dxa, na.action = na.omit)
Linear Hypotheses:
Estimate Std. Error z value Pr(>|z|)
(Intercept) == 0 -0.174273 0.020432 -8.53 < 2e-16
ageDecade == 0 -0.075306 0.006594 -11.42 < 2e-16
SexMen == 0 0.040435 0.008161 4.95 7.2e-07
RaceAfrAm == 0 0.058238 0.008069 7.22 5.3e-13
PovStatBelow == 0 0.019045 0.008296 2.30 0.02170
PhysBMI == 0 0.007394 0.000592 12.49 < 2e-16
MedHxAlcCurrYes == 0 0.005324 0.009265 0.57 0.56553
MedHxCigaretteCurrYes == 0 -0.011747 0.008217 -1.43 0.15284
ageDecade:SexMen == 0 0.027652 0.007636 3.62 0.00029
ageDecade:RaceAfrAm == 0 0.001540 0.007635 0.20 0.84014
(Univariate p values reported)
