I used a linear mixed effect model (taking into account repeated visits by some individuals) to look at neuropsych test performance and resting state data as a function of HIV status and marijuana use. If individuals tested positive for THC in their tox screen (I used thc_1 rather than thc_use_1) in at least one visit, I flagged them as a marijuana user. Unfortunately we do not currently have KMSK data for marijuana use, so I could not estimate quantity/length of use.

Previous lit has found some volumetric effects in in THC users in both HIV+ and HIV- individuals. Hippocampal volume changes have been seen in long term users of THC. Neuropsych tests have shown impaired memory, and especially verbal learning & memory, to be the strongest area affected by THC. There have been mixed reviews w/r/t THC and executive function, but the sample sizes in the studies that I’ve seen have been pretty small.

First of all, the proportion of THC use is higher in our HIV+ cohort than our controls. This aligns with expectations. Results of the chi-square test are below:

## 
##  Pearson's Chi-squared test with Yates' continuity correction
## 
## data:  final$hiv_status and final$thc___1
## X-squared = 16.652, df = 1, p-value = 4.49e-05

Looking at demographics, we can see that there are some differences between groups.

## [1] "Participant Age"
## # A tibble: 4 x 6
##   Cluster   count  mean    sd median   IQR
##   <fct>     <int> <dbl> <dbl>  <dbl> <dbl>
## 1 HIV-/THC-   190  49.1  14.1   48.5  18.8
## 2 HIV-/THC+    35  45.2  14.4   43.5  22.8
## 3 HIV+/THC-   299  51.6  11.4   54    14.5
## 4 HIV+/THC+   171  51.7  12.8   53    22
## 
##  Kruskal-Wallis rank sum test
## 
## data:  participant_age by Cluster
## Kruskal-Wallis chi-squared = 6.5383, df = 3, p-value = 0.08817
## [1] "Education"
## # A tibble: 4 x 6
##   Cluster   count  mean    sd median   IQR
##   <fct>     <int> <dbl> <dbl>  <dbl> <dbl>
## 1 HIV-/THC-    58  14.4  2.17   14    3   
## 2 HIV-/THC+    32  13.7  2.28   13.5  4   
## 3 HIV+/THC-    76  13.7  2.99   13    4   
## 4 HIV+/THC+   130  12.9  2.79   12.5  4.75
## 
##  Kruskal-Wallis rank sum test
## 
## data:  education by Cluster
## Kruskal-Wallis chi-squared = 12.743, df = 3, p-value = 0.005226
## [1] "KMSK Tobacco Scores"
## # A tibble: 4 x 6
##   Cluster   count  mean    sd median   IQR
##   <fct>     <int> <dbl> <dbl>  <dbl> <dbl>
## 1 HIV-/THC-    42  52.3 14.4    54.5  20.8
## 2 HIV-/THC+    29  44.4 14.7    43    23  
## 3 HIV+/THC-    43  53.8  8.72   55     7.5
## 4 HIV+/THC+    85  54.2 11.9    56    16
## 
##  Kruskal-Wallis rank sum test
## 
## data:  KMSK_Tobacco_TotalScore by Cluster
## Kruskal-Wallis chi-squared = 4.4174, df = 3, p-value = 0.2198
## [1] "Proportion of Male Participants"
## 
##  4-sample test for equality of proportions without continuity
##  correction
## 
## data:  case.vector out of total.vector
## X-squared = 11.036, df = 3, p-value = 0.01153
## alternative hypothesis: two.sided
## sample estimates:
##    prop 1    prop 2    prop 3    prop 4 
## 0.4655172 0.5312500 0.7368421 0.6136364
## [1] "Proportion of Depressed Participants"
## 
##  4-sample test for equality of proportions without continuity
##  correction
## 
## data:  case.vector out of total.vector
## X-squared = 3.878, df = 3, p-value = 0.2749
## alternative hypothesis: two.sided
## sample estimates:
##     prop 1     prop 2     prop 3     prop 4 
## 0.03906250 0.00000000 0.07262570 0.06993007

With my mixed effect model, I corrected for sex and age since those were different between cohorts, and took into account repeated measures for individuals. The following results show anything that was significant at the level of 0.01.

It seems like HIV+ individuals see effects to their functional connectivity based on THC use that HIV- individuals do not. There was also a drop off in performance on the Letter Number Sequencing test for HIV+ THC users as opposed to non-users.

There is a significant difference in Executive Function and on the Stroop test (CWIT3_Z) between HIV+ and HIV- individuals for both THC users and non-users, which is consistent with existing literature.

For THC non-users, there were differences based on HIV status in 2 functional domains: language and motor-psychomotor. This difference did not hold true for THC users. Instead, it looks like HIV- THC users look a lot like HIV+ individuals.

Then I put all of the significant parameters into a correlation matrix to see how they related to each other. Since we’re interested in changes that take place as a result of THC use, I only used participants who did not test positive for THC in that visit when looking for correlations. I used both HIV+ and HIV- participants.