Table 3.

Fixed-effect Coefficients From the Best-fitting Model Predicting Weight by Time to Death for 800 Male Decedents From the Baltimore Longitudinal Study of Aging, Maryland, 1958–2005^a

Age Group, Years	No.	Inflection Points, Years				Weight at Death		Weight Loss Rate After Inflection		Change in Weight Loss Rate Before Inflection 1		Change in Weight Loss Rate Before Inflection 2
		k1	CIb	k2	CIb	B0	P Value	B1 (Time)	P Value	B2 (Time > k1)	P Value	B3 (Time > k2)	P Value
Average Model for All Age Groups, All Causes of Death (N = 800)
60–107	800	9	7, 10	17	16, 18	73.517	<0.001	0.390	<0.001	−0.270	<0.001	−0.109	0.019
Model With Varied Knot and Slopes by Age Group, All Causes of Death (N = 800)
60–69	87	3	1, 6	15	8, 19	77.850	<0.001	1.011	0.001	−1.123	<0.001	−0.063	0.618
70–79	211	7	6, 9	18	17, 21	75.781	<0.001	0.566	<0.001	−0.463	0.002	−0.250	0.012
80–89	320	3	2, 4	16	14, 17	70.903	<0.001	1.263	<0.001	−1.003	0.001	−0.189	0.003
≥90	182	10	9, 12	16	14, 19	68.096	<0.001	0.468	<0.001	−0.267	0.036	−0.099	0.256
Model With Varied Knot and Slopes by Age Group, Cancer Deaths (N = 143)
60–79	71	3	2c, 4	23	21, 23c	75.496	<0.001	0.874	0.032	−0.955	0.024	−0.099	0.439
≥80	72	3	2c, 5	18	16, 22	72.581	<0.001	1.424	0.004	−1.337	0.009	0.062	0.551
Model With Varied Knot and Slopes by Age Group, Cardiovascular Deaths (N = 349)
60–69	47	5	5c, 5c	24	20, 25c	76.043	<0.001	0.547	0.017	−0.597	0.020	0.017	0.969
70–79	94	8	6, 9	20	18, 23	76.157	<0.001	0.447	0.003	−0.365	0.124	−0.222	0.251
80–89	133	10	8, 12	25	22, 25d	71.755	<0.001	0.507	<0.001	−0.374	0.001	−0.233	0.054
≥90	75	9	7, 11	18	14, 20	66.061	<0.001	0.797	<0.001	−0.577	0.002	−0.123	0.460
Model With Varied Knot and Slopes by Age Group, Other Deaths (N = 211)
60–79	63	6	4, 10	17	15, 20	72.230	<0.001	1.192	<0.001	−1.066	0.001	−0.346	0.041
80–89	87	4	4c, 5	16	14, 18	68.992	<0.001	0.870	0.028	−0.469	0.290	−0.292	0.044
≥90	61	4	4c, 7	10	8, 11	65.655	<0.001	0.430	0.282	0.244	0.636	−0.467	0.024

Abbreviation: CI, confidence interval.

^aModel includes no intercept term, so that coefficients can be interpreted directly for each age group. P values represent differences from zero based on 2-sided t test.

^bConfidence intervals are >95%. Knot placements were estimated by using a likelihood ratio testing approach, with the last integer valued knots that were significantly different from the best-fitting knot at the 0.025 level on either side included as upper and lower confidence limits.

^cIndicates that the model would not converge at the next value beyond this knot point.

^dIndicates outer limit of tested knots (25 years before death).