interpretation of longitudinal data

[Kyra]

Hi,

I am working with longitudinal data for the first time. I need a little help with analysis.

I have lab values preop, 6 months, 12 months, 24 months postop for creatinine for a group of patients.

I want to compare the trajectory between obese and non obese group.

part of my dataset is:

DMRN	obesity	time	creatinine
7300031	nonob	0	0.8
7321580	nonob	0	0.93
7281606	nonob	0	0.73
7225922	nonob	0	0.68
7300031	nonob	6	1.79
7321580	nonob	6
7281606	nonob	6
7225922	nonob	6	1.21
7300031	nonob	12	1.6
7321580	nonob	12
7281606	nonob	12
7225922	nonob	12	1.12
7300031	nonob	24	1.53
7321580	nonob	24	1.2
7281606	nonob	24	1.4
7225922	nonob	24	1.06

 i have used the below code.

proc mixed data=red.new method=ml;
class DMRN obesity ;
model creatinine = obesity time obesity*time / s;
repeated / type=ar(1) subject=DMRN group=obesity;
run;

I am pasting my output

Model Information
Data Set	RED.NEW
Dependent Variable	creatinine
Covariance Structure	Autoregressive
Subject Effect	DMRN
Group Effect	obesity
Estimation Method	ML
Residual Variance Method	None
Fixed Effects SE Method	Model-Based
Degrees of Freedom Method	Between-Within


Class Level Information
Class	Levels	Values
DMRN	74	2290228 2480811 2825809 3143067 3145614 3430846 3597837 6021584 6305161 7157755 7163265 7225922 7244425 7265190 7281606 7286918 7290563 7300031 7312662 7320783 7321580 7321866 7325162 7330013 7332704 7332787 7336014 7340994 7341213 7341980 7342354 7345560 7349764 7361060 7363079 7364229 7369669 7372753 7375796 7376577 7377069 7377576 7382679 7382681 7391466 7395193 7395311 7399264 7410841 7410870 7412002 7412512 7412822 7412865 7412900 7419006 7434224 7446572 7448861 7455344 7478866 7483328 7483731 7490221 7502069 7502324 7524993 7526872 7609175 7633993 7675067 7676515 7682442 7718968
obesity	2	nonob obese


Dimensions
Covariance Parameters	4
Columns in X	6
Columns in Z	0
Subjects	74
Max Obs per Subject	4


Number of Observations
Number of Observations Read	298
Number of Observations Used	228
Number of Observations Not Used	70


Iteration History
Iteration	Evaluations	-2 Log Like	Criterion
0	1	-6.21039560	
1	2	-33.31404693	0.03275555
2	1	-41.35741962	0.00452043
3	1	-42.53522587	0.00029572
4	1	-42.60745297	0.00000269
5	1	-42.60807686	0.00000000


Convergence criteria met.


Covariance Parameter Estimates
Cov Parm	Subject	Group	Estimate
Variance	DMRN	obesity nonob	0.04987
AR(1)	DMRN	obesity nonob	0.3509
Variance	DMRN	obesity obese	0.06532
AR(1)	DMRN	obesity obese	0.5371


Fit Statistics
-2 Log Likelihood	-42.6
AIC (Smaller is Better)	-26.6
AICC (Smaller is Better)	-26.0
BIC (Smaller is Better)	-8.2


Null Model Likelihood Ratio Test
DF	Chi-Square	Pr > ChiSq
3	36.40	<.0001


Solution for Fixed Effects
Effect	obesity	Estimate	Standard
Error	DF	t Value	Pr > |t|
Intercept		1.1036	0.04566	72	24.17	<.0001
obesity	nonob	-0.09035	0.05476	72	-1.65	0.1033
obesity	obese	0	.	.	.	.
time		0.006546	0.002878	152	2.27	0.0243
time*obesity	nonob	0.002397	0.003590	152	0.67	0.5054
time*obesity	obese	0	.	.	.	.


Type 3 Tests of Fixed Effects
Effect	Num DF	Den DF	F Value	Pr > F
obesity	1	72	2.72	0.1033
time	1	152	18.61	<.0001
time*obesity	1	152	0.45	0.5054

Please help me interpret.

Is the trend of creatinine not significantly different among obese and non obese (p-value- 0.1033)?

Thanks.

[Kyra]

I am attaching another output where i have used time as categorical variable.

Model Information
Data Set	RED.NEW
Dependent Variable	creatinine
Covariance Structure	Autoregressive
Subject Effect	DMRN
Group Effect	obesity
Estimation Method	ML
Residual Variance Method	None
Fixed Effects SE Method	Model-Based
Degrees of Freedom Method	Between-Within


Class Level Information
Class	Levels	Values
DMRN	74	2290228 2480811 2825809 3143067 3145614 3430846 3597837 6021584 6305161 7157755 7163265 7225922 7244425 7265190 7281606 7286918 7290563 7300031 7312662 7320783 7321580 7321866 7325162 7330013 7332704 7332787 7336014 7340994 7341213 7341980 7342354 7345560 7349764 7361060 7363079 7364229 7369669 7372753 7375796 7376577 7377069 7377576 7382679 7382681 7391466 7395193 7395311 7399264 7410841 7410870 7412002 7412512 7412822 7412865 7412900 7419006 7434224 7446572 7448861 7455344 7478866 7483328 7483731 7490221 7502069 7502324 7524993 7526872 7609175 7633993 7675067 7676515 7682442 7718968
obesity	2	nonob obese
time	4	0 6 12 24


Dimensions
Covariance Parameters	4
Columns in X	15
Columns in Z	0
Subjects	74
Max Obs per Subject	4


Number of Observations
Number of Observations Read	298
Number of Observations Used	228
Number of Observations Not Used	70


Iteration History
Iteration	Evaluations	-2 Log Like	Criterion
0	1	-57.19407331	
1	2	-108.82313308	0.08071153
2	1	-120.21075647	0.02760509
3	1	-129.01528882	0.00349022
4	1	-130.06362297	0.00009803
5	1	-130.09116079	0.00000014
6	1	-130.09119954	0.00000000


Convergence criteria met.


Covariance Parameter Estimates
Cov Parm	Subject	Group	Estimate
Variance	DMRN	obesity nonob	0.03663
AR(1)	DMRN	obesity nonob	0.5359
Variance	DMRN	obesity obese	0.05594
AR(1)	DMRN	obesity obese	0.6589


Fit Statistics
-2 Log Likelihood	-130.1
AIC (Smaller is Better)	-106.1
AICC (Smaller is Better)	-104.6
BIC (Smaller is Better)	-78.4


Null Model Likelihood Ratio Test
DF	Chi-Square	Pr > ChiSq
3	72.90	<.0001


Solution for Fixed Effects
Effect	obesity	time	Estimate	Standard
Error	DF	t Value	Pr > |t|
Intercept			1.2085	0.05106	72	23.67	<.0001
obesity	nonob		-0.03980	0.06160	72	-0.65	0.5202
obesity	obese		0	.	.	.	.
time		0	-0.1706	0.05884	148	-2.90	0.0043
time		6	0.08799	0.05577	148	1.58	0.1168
time		12	0.04262	0.04608	148	0.92	0.3565
time		24	0	.	.	.	.
obesity*time	nonob	0	-0.07045	0.07211	148	-0.98	0.3302
obesity*time	nonob	6	-0.01492	0.06949	148	-0.21	0.8303
obesity*time	nonob	12	0.01941	0.05773	148	0.34	0.7372
obesity*time	nonob	24	0	.	.	.	.
obesity*time	obese	0	0	.	.	.	.
obesity*time	obese	6	0	.	.	.	.
obesity*time	obese	12	0	.	.	.	.
obesity*time	obese	24	0	.	.	.	.


Type 3 Tests of Fixed Effects
Effect	Num DF	Den DF	F Value	Pr > F
obesity	1	72	1.68	0.1990
time	3	148	43.45	<.0001
obesity*time	3	148	0.72	0.5430

I am attaching my  sas code too.

proc mixed data=red.new method=ml;
   class DMRN obesity time;
   model creatinine = obesity time obesity*time / s;
   repeated / type=ar(1) subject=DMRN group=obesity;
run;

Please let me know if this makes more sense. And please let me know how to interpret this.