Table 4.

Association between baseline PINP, ICTP and change in femoral neck BMD

	Symbol	Estimate ± SE	P-value
Analysis marker: sPINP
Estimated parameters
Intercept (baseline)	α	943 ± 13	<0.0001
Rate of change	β	-3.5 ± 1.0	0.0004
sPINP	γ	0.7 ± 1.1	0.4953
Variance components
Baseline	${\sigma }_{\alpha }^2$	16661 ± 2454	<0.001
Rate of change	${\sigma }_{\beta }^2$	44 ± 16	0.004
Covariance	${\sigma }_{\alpha \beta }^2$	54 ± 131	0.692
Random error	${\sigma }_e^2$	902 ± 130	<0.001
Analysis marker: sICTP
Estimated parameters
Intercept (baseline)	α	943 ± 13	<0.0001
Rate of change	β	-3.6 ± 1.0	0.0008
sICTP	γ	-0.6 ± 1.3	0.629
Variance components
Baseline	${\sigma }_{\alpha }^2$	16643 ± 2453	<0.0001
Rate of change	${\sigma }_{\beta }^2$	43 ± 16	0.0041
Covariance	${\sigma }_{\alpha \beta }^2$	56 ± 139	0.6848
Random error	${\sigma }_e^2$	907 ± 131	<0.0001

sICTP, serum C-terminal telopeptide of type I collagen; sPINP, serum N-terminal propeptide of type I collagen; BMD bone mineral density.

Note of interpretation: Results in part (a) suggest that the BMD measurement (mg/cm²) for a subject at time t (denoted by BMD_t) can be represented by the equation: BMD_t = 943 – 3.5*Year – 9.2* sPINP0, where "Year" is 0, 1, 2, ..., and sPINP0 is the measurement of sPINP at baseline for the subject. In other words, femoral neck BMD was decreased by -3.5 mg/cm² per year. There was no significant association between baseline sPINP and change in BMD (p = 0.4953). The interpretation of variance components is similar to that in Table 3.