Created 2013-01-02; last modified Sun Jan 6 11:44:17 2013
We have an interesting, if perhaps inexplicable finding. Now let's see if it survives a "stress" test by adding some additional covariates.
So far we've found a significant 3-way interaction of BDNF, CRP, and race (Table 1.1). The interesting feature of this result is that there are significant race differences at elevated levels of CRP such that when BDNF is high Oxo-G is low in Blacks and Oxo-G is high in whites. This relationship is shown in Figure 1, wherein the relationship is expressed a varying levels of CRP (the differences between Blacks and whites increasingly "fans out" as CRP increases).
Table 1.1. Association of BDNF, CRP, race, and 3-way interaction with Oxo
| B | RAfrAm | C | B:RAfrAm | B:C | RAfrAm:C | B:RAfrAm:C | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Oxo | -3.3e-07 | -0.031 | * | 0.0018 | 1.5e-06 | *** | 1.4e-07 | 0.008 | -4.4e-07 | * |
Now let's do a "covariate stress test" to see if some other covariates make this relationship disappear. In order, I added cigarette smoking (Table 1.2), any current substance abuse (Table 1.3), diabetes diagnosis (Table 1.4), and presence of metabolic syndrome (Table 1.5).
Table 1.2. Association of cigarette smoking, BDNF, CRP, race, and 3-way interaction with Oxo
| B | RAfrAm | C | MedHxCigaretteCurrYes | B:RAfrAm | B:C | RAfrAm:C | B:MedHxCigaretteCurrYes | B:RAfrAm:C | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Oxo | 3.5e-08 | -0.026 | 0.0047 | 0.0024 | 1.5e-06 | ** | 4.1e-08 | 0.007 | -2.9e-07 | -4.4e-07 | * |
Table 1.3. Association of substance abuse, BDNF, CRP, race, and 3-way interaction with Oxo
| B | RAfrAm | C | subAbuseTRUE | B:RAfrAm | B:C | RAfrAm:C | B:subAbuseTRUE | B:RAfrAm:C | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Oxo | -3e-07 | -0.032 | * | 0.0011 | 0.00086 | 1.5e-06 | *** | 1.8e-07 | 0.0076 | -2.5e-07 | -4.3e-07 | * |
Table 1.4. Association of diabetes diagnosis, BDNF, CRP, race, and 3-way interaction with Oxo
| B | RAfrAm | C | diabetesYes | B:RAfrAm | B:C | RAfrAm:C | B:diabetesYes | B:RAfrAm:C | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Oxo | -3.3e-07 | -0.03 | * | 0.0032 | -0.0098 | 1.5e-06 | *** | 1.3e-07 | 0.007 | 9.4e-08 | -4.2e-07 | * |
Table 1.5. Association of metabolic syndrome, BDNF, CRP, race, and 3-way interaction with Oxo
| B | RAfrAm | C | MSatpYes | B:RAfrAm | B:C | RAfrAm:C | B:MSatpYes | B:RAfrAm:C | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Oxo | -7.1e-07 | -0.04 | ** | 0.0031 | -0.009 | 1.9e-06 | *** | 3.8e-08 | 0.0087 | 8.7e-07 | * | -4.5e-07 |
The 3-way interaction survives the "stress" test after adding cigarette smoking, substance abuse, and diabetes diagnosis. The 3-way interaction does not survive when metabolic syndrome is added (Table 1.5). This is likely due to the presence of blood pressure in the definition for metabolic syndrome, not the presence of diabetes-like conditions because diabetes itself has no effect on the 3-way interaction (Table 1.4).