Introduction

Neurofeedback(NFB) is a type of Biofeedback(BFB) that is used to help individuals regulate their brainwaves and improve their cognitive functioning. NFB has been used to treat a variety of conditions, including anxiety, depression, attention-deficit/hyperactivity disorder (ADHD), and post-traumatic stress disorder (PTSD).

This dataset includes gender, age, and reported values of symptoms for 67 veterans during a NFB therapy study. The study design involves 20 sessions of NFB, but participants were allowed to complete the study early if treatment resulted in a full alleviation of symptoms.

  • n = 67
  • Age = Age
  • Gender = Male/Female
  • Session = First/Last – also called Treatment(Tx.)
  • Symptom = Activity, Anger, Anxiety, Clouding, Depression, Fatigue, Pain, Sleep
  • PTSD = Total score for PCL-M starting
  • PTSD Final = Total score for PCL-M finishing

Methods

The forms below (Current Symptom Ratings and PCL-M) were filled out throughout the study by participants.

Current Symptoms

Current Symptoms Rating measures at broad wellness.

PTSD PCL-M

PCL-M measures PTSD symptoms for Veterans.

Participant Demographics

Participants included 67 total veterans (14 Female, 53 Male), of Age = 38.64 ± 8.87.
Participant ages range from 23 to 63, with an approximately normal distribution.

Age

Solid line: Age Density Curve
Segmented Line: Normal Curve

Gender

Roughly 4 times as many male veterans to female veterans.

Pre and Post Treatment Analysis

Current Symptoms Change

Symptom ratings shifted from mostly 6-8s to mostly 1-3s as seen in the bar and density plots.
Significant improvements for each measure can be seen from the box plots by symptom category.

Plots

Pre/Post Barplot

Density

Total Score Boxplots

Paired T-Test

Comparing the total value of participants symptom ratings:
A statistically significant improvement in symptom ratings from first treatment to last treatment.
*Mean differences 21.87, 95% conf. int. = 18.69

## 
##  Paired t-test
## 
## data:  PairedTx$Tx.First and PairedTx$Tx.Last
## t = 11.465, df = 66, p-value < 2.2e-16
## alternative hypothesis: true difference in means is greater than 0
## 95 percent confidence interval:
##  18.68388      Inf
## sample estimates:
## mean of the differences 
##                21.86567

PTSD Change

Plots

Paired T-Test

Comparing the value of participants PTSD PCL-M scores:
A statistically significant improvement in PCL-M ratings from first treatment to last treatment.
*Mean differences 17.46, 95% conf. int. = 14.2

## 
##  Paired t-test
## 
## data:  df6$PTSD and df6$PTSD_FINAL
## t = 8.9171, df = 66, p-value = 3.052e-13
## alternative hypothesis: true difference in means is greater than 0
## 95 percent confidence interval:
##  14.19566      Inf
## sample estimates:
## mean of the differences 
##                17.46269

20 Sessions Analysis

54 of the 67 participants completed all 20 sessions. The following visuals and models will be analyzing those 54.

Symptom Category Changes over Sessions

All Symptoms

Activity

Anxiety

Anger

Clouding

Depression

Fatigue

Pain

Sleep

Visual Symptom Rating Progressions

The visuals below provide an overview of the change in symptoms over the course of treatment.

Heatmap

Distribution Change over Treatments

Individual Current Symptom Ratings of participants by Session

Distribution Total Score Change over Treatments

Sum totals for each Current Symptom Ratings of participants by Session

Linear Model

Looking at the relationship between sessions of neurofeedback on total ratings for the Current Symptoms Ratings.

Linear Regression of Symptom Total over Treatment

Significance: p-value <2e-16 for Treatment and Model. Low R-squared indicates model does not account for ~90% variance.
Treatment estimate is about -1 total rating per session. For those who complete 20 sessions, this model suggests a decrease of about 2.5 for each symptom rating.

## 
## Call:
## lm(formula = Value ~ Treatment, data = df20)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -30.644 -11.774  -2.735  10.291  53.942 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 40.59251    1.00962   40.21   <2e-16 ***
## Treatment   -0.97413    0.08428  -11.56   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 15.97 on 1078 degrees of freedom
## Multiple R-squared:  0.1103, Adjusted R-squared:  0.1094 
## F-statistic: 133.6 on 1 and 1078 DF,  p-value: < 2.2e-16
## Analysis of Variance Table
## 
## Response: Value
##             Df Sum Sq Mean Sq F value    Pr(>F)    
## Treatment    1  34076   34076  133.59 < 2.2e-16 ***
## Residuals 1078 274977     255                      
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Conf. Interval Plot

The fit line and the confidence interval show the trend of roughly 1 total rating improvement per session.
The low R-squared can be seen by the wide prediction interval.

Boxplots of Scores over Sessions

High scoring outliers for the later sessions can be seen in the top right.

Conclusion

NFB treatment showed to significantly improve overall wellness symptoms and PTSD symptoms. From the paired t-test comparisons, we can say with 95% confidence that the true mean difference from beginning NFB and ending NFB treatment is at least 18.68 points for the Current Symptom Ratings and at least 14.2 for the PCL-M scores. VA standards for clinically meaningful treatment are a 10-20 point change.
See (https://www.ptsd.va.gov/professional/assessment/documents/PCL_handoutDSM4.pdf) for more details.

From the linear model, the weak R-squared and the high residual standard error suggest this model may not be a good fit. This could be due to the sample size and the inclusion of outliers. The sample improved at a rate of about 1 total Current Symptom Rating per session. For future studies, larger sample sizes would be able to provide a more accurate predictive model.