Created 2012-12-22; last modified Sun Dec 30 12:09:27 2012

Distribution of BDNF & log(BDNF)

It's apparent that BDNF is skewed (Figure 1). Log transformation appears to make it bimodel, not a great improvement. We can look at both, though I suspect that the untransformed is the form to use.

Figure 1. Distributions of BDNF and log(BDNF) in complete sample

The pattern of distributions are different within CRP groups and different for BDNF untransformed and transformed (Figure 2). There is no clear pattern of differences among groups.

Figure 2. Distributions of BDNF and log(BDNF) within CRP groups

Association of Oxo with BDNF & log(BDNF)

Figure 3. Association of Oxo with BDNF & log(BDNF) in complete sample

Figure 4. Association of Oxo with BDNF & log(BDNF) within CRP groups

Associations with BDNF & log(BDNF)

Complete sample

Correlations of BDNF and log(BDNF) with CRP, BMI, blood pressure, pulse pressure, total cholesterol, eGFR, and some inflammatory markers for complete sample

        Oxo      CRPlog10 PhysBMI  PhysBPsitDia PhysBPsitSys PP       LabResChol eGFR     hIL18    hTNFa    hRAGE   
hBDNF    0.17     0.10    -0.01    -0.15         0.13         0.28**  -0.04      -0.06     0.02    -0.02     0.48***
logBDNF  0.04     0.03    -0.05    -0.19*        0.04         0.19*   -0.18      -0.07    -0.06     0.07     0.53***

n
        Oxo CRPlog10 PhysBMI PhysBPsitDia PhysBPsitSys  PP LabResChol eGFR hIL18 hTNFa hRAGE
hBDNF   115      115     112          109          109 109        115  115   115    93   115
logBDNF 115      115     112          109          109 109        115  115   115    93   115

Figure 5. Scatterplot matrix for Oxo, BDNF, and other measures in complete sample

Within CRP groups

Correlations of BDNF and log(BDNF) with CRP, BMI, blood pressure, pulse pressure, total cholesterol, eGFR, and some inflammatory markers by CRP group.

oxo$CRPgroup: Lo
        Oxo      CRPlog10 PhysBMI  PhysBPsitDia PhysBPsitSys PP       LabResChol eGFR     hIL18    hTNFa    hRAGE   
hBDNF    0.36*    0.04     0.03     0.04         0.33*        0.44**   0.13      -0.20     0.02    -0.07     0.02   
logBDNF  0.07    -0.01     0.06    -0.10         0.15         0.31     0.03      -0.21    -0.15     0.02    -0.21   

n
        Oxo CRPlog10 PhysBMI PhysBPsitDia PhysBPsitSys PP LabResChol eGFR hIL18 hTNFa hRAGE
hBDNF    37       37      37           35           35 35         37   37    37    37    37
logBDNF  37       37      37           35           35 35         37   37    37    37    37
------------------------------------------------------------------------------------------------------------------------------------------------------ 
oxo$CRPgroup: Mid
        Oxo      CRPlog10 PhysBMI  PhysBPsitDia PhysBPsitSys PP       LabResChol eGFR     hIL18    hTNFa    hRAGE   
hBDNF   -0.00     0.01    -0.19    -0.09         0.16         0.28    -0.24      -0.15     0.24     0.52*    0.58***
logBDNF  0.03     0.17    -0.11    -0.04         0.21         0.29    -0.27      -0.23     0.31     0.40     0.45** 

n
        Oxo CRPlog10 PhysBMI PhysBPsitDia PhysBPsitSys PP LabResChol eGFR hIL18 hTNFa hRAGE
hBDNF    39       39      39           38           38 38         39   39    39    17    39
logBDNF  39       39      39           38           38 38         39   39    39    17    39
------------------------------------------------------------------------------------------------------------------------------------------------------ 
oxo$CRPgroup: Hi
        Oxo      CRPlog10 PhysBMI  PhysBPsitDia PhysBPsitSys PP       LabResChol eGFR     hIL18    hTNFa    hRAGE   
hBDNF    0.14    -0.03    -0.01    -0.17         0.07         0.19     0.16      -0.03     0.14    -0.05     0.37*  
logBDNF  0.13     0.04    -0.04    -0.19         0.08         0.21     0.07      -0.08     0.12    -0.06     0.42** 

n
        Oxo CRPlog10 PhysBMI PhysBPsitDia PhysBPsitSys PP LabResChol eGFR hIL18 hTNFa hRAGE
hBDNF    39       39      36           36           36 36         39   39    39    39    39
logBDNF  39       39      36           36           36 36         39   39    39    39    39

Associations of BDNF with Oxo adjusting for covariates

Adjusting for demographics

First examine association of BDNF (B) with Oxo adjusting for race & poverty status with interactions.

There is no significant 3-way interaction of BDNF by race by poverty status on Oxo (Table 1.1), but there is a significant 2-way interaction between BDNF & race (Table 1.2). We can drop poverty status because it is not associated significantly with Oxo in these analyses. Although the effect size for the significant 2-way interaction between BDNF & race is very small (Table 1.3), it is significant in the presence of the main effects (main effects must be retained for significant interactions).

Table 1.1. Association of BDNF, race, poverty status with 3-way interaction on Oxo

	B		RAfrAm		PBelow		B:RAfrAm		B:PBelow		RAfrAm:PBelow		B:RAfrAm:PBelow
Oxo	-0.00000005		-0.0045		0.0078		0.00000014		-0.00000016		-0.027		0.00000095

Table 1.2. Association of BDNF, race, poverty status with 2-way interactions on Oxo

	B		RAfrAm		PBelow		B:RAfrAm		B:PBelow		RAfrAm:PBelow
Oxo	-0.00000018		-0.014		-0.003		0.00000053	*	0.00000032		-0.0034

Table 1.3. Association of BDNF, race, & 2-way interaction on Oxo

	B		RAfrAm		B:RAfrAm
Oxo	-0.000000075		-0.018		0.00000064	*

Adjusting for race, blood pressure, & inflammatory markers

Second, add blood pressure & inflammatory markers with interactions to test whether other covariates eliminate the association between BDNF & Oxo.

Start by adding log(CRP) (C): there is a significant 3-way interaction of BDNF by race by log(CRP) on Oxo (Table 2.1).

Table 2.1. Association of BDNF, race, & log(CRP) on Oxo

	B		RAfrAm		C		B:RAfrAm		B:C		RAfrAm:C		B:RAfrAm:C
Oxo	-0.00000033		-0.031	*	0.0018		0.0000015	***	0.00000014		0.008		-0.00000044	*

CRP is significantly associated with Oxo after adding other inflammatory markers, none of which were significant (Table 3.1).

Table 3.1. Association of BDNF, race, log(CRP), & other inflammatory markers on Oxo

	B		RAfrAm		C		hIL18		hTNFa		hRAGE		B:RAfrAm		B:C		RAfrAm:C		B:RAfrAm:C
Oxo	-0.00000025		-0.031		0.0025		0.000013		0.000023		0.0000013		0.0000015	**	0.00000013		0.0083		-0.00000045	*

BMI, total cholesterol, & eGFR are not significant covariates (Table 4.1).

Table 4.1. Association of BDNF, race, log(CRP), BMI, total cholesterol, & eGFR on Oxo

	B		RAfrAm		C		BMI		LabResChol		eGFR		B:RAfrAm		B:C		RAfrAm:C		B:RAfrAm:C
Oxo	-0.00000031		-0.037	**	-0.001		0.00055		0.00003		0.00012		0.0000017	***	0.00000018		0.012		-0.0000006	**

Systolic blood pressure is a significant covariate; diastolic blood pressure is not (Table 5.1). However, the 3-way interaction (BDNF by race by log(CRP)) is nonsignificant after adding blood pressure (Table 5.2). Systolic blood pressure (Table 5.3) or pulse pressure (Table 5.4) are significant covariates by themselves (but not in the same analysis [not shown]).

Table 5.1. Association of BDNF, race, log(CRP), & diastolic & systolic blood pressure on Oxo

	B		RAfrAm		C		BPdia		BPsys		B:RAfrAm		B:C		RAfrAm:C		B:RAfrAm:C
Oxo	-0.00000043		-0.033	*	0.00058		-0.00031		0.00064	*	0.0000012	**	0.00000016		0.0061		-0.00000032

Table 5.2. Association of BDNF, race, log(CRP), & systolic blood pressure on Oxo

	B		RAfrAm		C		BPsys		B:RAfrAm		B:C		RAfrAm:C		B:RAfrAm:C
Oxo	-0.00000038		-0.032	*	0.0011		0.00051	**	0.0000012	**	0.00000015		0.0059		-0.00000033

Table 5.3. Association of BDNF, race, & systolic blood pressure on Oxo

	B		RAfrAm		BPsys		B:RAfrAm
Oxo	-0.0000002		-0.024	*	0.0006	***	0.0000007	**

Table 5.4. Association of BDNF, race, & pulse pressure on Oxo

	B		RAfrAm		PP		B:RAfrAm
Oxo	-0.00000025		-0.025	**	0.00082	***	0.00000068	**

The interaction between BDNF & race survives regardless of which covariates we apply. It seems we have a choice between presenting significant adjustment for an inflammatory marker (CRP) or significant adjustment for an arterial marker (pulse pressure, a measure of arterial stiffness, or systolic blood pressure). The significant association of BDNF with Oxo is interesting (Table 2.1). It suggests that whites and Blacks DNA repair diverges as CRP increases (Figure 6). The significant association of BDNF with Oxo after adjusting for pulse pressure (Figure 7) builds on the previous finding but may be too similar for a novel result in a good journal (Table 5.4). The main effect of pulse pressure is seen in the change of levels but not change in interactions as PP increases. However, Oxo diverges in Blacks and whites at all levels of PP.

Figure 6. Oxo on BDNF by 3 levels of CRP

Figure 7. Oxo on BDNF by 3 levels of PP