Trulli
source

Overview

In this case study, we will look at the results of a clinical trial exploring the effectiveness of a new medication called dimarta on reducing histamine in patients with a disease that leads to chronically high histamine levels. In the study, 300 patients were assigned to one of three different treatment arms. One arm was given a placebo. The other arm was given adiclax – the standard of care for the disease. Finally, the third arm was given dimarta. There were two main measures of interest in the trial: patient’s changes in histamine from the beginning to the end of the trial, and their change in quality of life (measured by self report).

In addition to exploring the effects of the three medications, the researchers are interested in the extent to which three different biomarkers, dw, ms, and np, are correlated with therapeutic outcomes. In other words, to patients that express one or more of these biomarkers have better, or worse, outcomes that those that do not express these biomarkers?

Datasets

File Rows Columns Description
dimarta_trial.csv 300 6 Key DIMARTA trial outcomes
dimarta_biomarker.csv 900 3 Biomarker status’ for 3 different biomarkers for each patient.
dimarta_demographics.csv 300 5 Demographic information for each patient

Column Descriptions

dimarta_trial.csv

Table1. dimarta_trial.csv variable description:
Variable Description
PatientID Unique patient id
arm Treatment arm, either 1 = placebo, 2 = adiclax (the standard of treatment), or 3 = dimarta (the target drug)
histamine_start histamine value at the start of the trial
histamine_end histamine value at the end of the trial
qol_start Patient’s rated quality of life at the start of the trial
qol_end Patient’s rated quality of life at the end of the trial

dimarta_demographics.csv

Table2. dimarta_demographics.csv variable description:
Variable Description
PatientID Unique patient id
age Patient age
gender Patient gender, 0 = male, 1 = female
site Site where the clinical trial was conducted
diseasestatus Status of the patient’s disease at start of trial

dimarta_biomarker.csv

Table3. dimarta_biomarker.csv variable description:
Variable Description
PatientID Unique patient id
Biomarker One of three biomarkers: dw, ms, and np
BiomarkerStatus Result of the test for the biomarker.

Tasks

A - Getting Setup

  1. Open your baselrbootcamp R project. It should already have the folders 1_Data and 2_Code.

  2. Open a new R script and save it as a new file in your R folder called dimarta_casestudy.R. At the top of the script, using comments, write your name and the date. Then, load the tidyverse package. Here’s how the top of your script should look:

## My Name
## The Date
## Dimarta - Case Study

library(tidyverse)

B - Data I/O

  1. Using read_csv(), load the dimarta_trial.csv, dimarta_demographics.csv, and dimarta_biomarker.csv datasets as three new objects called trial_df, demographics_df, and biomarker_df.

  2. Get a first impression of the objects you just created by exploring them with a mixture of the View(), head(), names(), and str() functions. Were they all loaded correctly?

C - Data Wrangling

  1. Using rename(), change the name of the column arm in the trial_df data to StudyArm.

  2. Using the table() function, look at the values of the StudyArm column in trial_df. You’ll notice the values are 1, 2, and 3. Using mutate() and case_when() change these values to the appropriate names of the study arms (look at the variable descriptions to see which is which!)

  3. In the demographics_df data, you’ll see that gender is coded as 0 and 1. Using mutate() create a new column in demographics_df called gender_c that shows gender as a string, where 0 = “male”, and 1 = “female”.

  4. Now let’s create a new object called dimarta_df that combines data from trial_df and demographics_df. To do this, use left_join() to combine the trial_df data with the demographics_df data. This will merge the two datasets so you can have the study results and demographic data in the same dataframe. Make sure to assign the result to a new object called dimarta_df

# Create a new dataframe called dimarta_df that contains both trial_df and demographics_df
dimarta_df <- trial_df %>%
  left_join(demographics_df)
  1. You’ll notice that the biomarker_df dataframe is in the ‘long’ format, where each row is a patient’s biomarker result. Using the code below, making use of the spread() function, create a new dataframe called biomarker_wide_df where each row is a patient, and the results from different biomarkers are in different columns. When you finish, look at biomarker_wide_df to see how it looks!
# Convert biomarker_df to a wide format using spread()
biomarker_wide_df <- biomarker_df %>%
  spread(Biomarker, BiomarkerStatus)

biomarker_wide_df
  1. Now, using the left_join function, add the biomarker_wide_df data to the dimarta_df data! Now you should hve all of the data in a single dataframe called dimarta_df

  2. View dimarta_df to make sure the data look correct! The data should have one row for each patient, and 13 separate columns, including dw, ms, and np

  3. Using the mean() function, calculate the mean age of all patients.

  4. Using the following template, find out how many male and female patients were in the trial.

dimarta_df %>%
  group_by(XXX) %>%
  summarise(
    Counts = n()    
  )
  1. Now, using similar code, find out how many patients were assigned to each study arm.

  2. Find out how many men and women were assigned to each study arm (Hint: You can use very similar code to what you used above, just add a second grouping variable!)

  3. Using mutate(), add a new column to the dimarta_df data called histamine_change that shows the change in patient’s histamine levels from the start to the end of the trial (Hint: just subtract histamine_start from histamine_end!)

  4. Using mutate() again, add a new column to dimarta_df called qol_change that shows the change in patient’s quality of life.

  5. Calculate the percentage of patients who tested positive for each of the three biomarkers using the following template (Hint: If you calculate the mean() of a logical vector, you will get the percentage of TRUE values!)

# Calculate percent of patients with positive biomarkers

dimarta_df %>%
  summarise(
    dw_mean = mean(XXX),
    ms_percent = mean(XXX),
    np_percent = mean(XXX)
  )
  1. Were there different distributions of age in the different trial sites? To answer this, separately calculate the mean and standard deviations of patient ages in each site. (Hint: group the data by site, then calculate two separate summary statistics: age_mean = mean(age), and age_sd = sd(age).

  2. Calculate the mean change in histamine results separately for each study site

  3. Calculate the mean change in histamine results (histamine_change) for each study arm. Which study arm had a largest decrease in histamine?

  4. Calculate the mean change in quality of life (qol_change) for each study arm. Do the results match what you found with the histamine results?

D - Plotting

  1. Run the following code block to visualise the relationship between study arm and histamine change. What do you find?
ggplot(data = dimarta_df,
       mapping = aes(x = StudyArm, 
                     y = histamine_change)) +
  geom_boxplot()
  1. Create the same plot as above, but instead of analysing study arm, try analysing gender. What do you find? Did one gender have better histamine improvements than the other?

E - Statistics

  1. Using t.test() conduct a t-test comparing the change in histamine results between the placebo and dimarta. Did dimarta differ from the placebo?
# T.test comparing change in histamine between placebo and dimarta

t.test(formula = XX ~ XX,
       data = XXX %>%
              filter(XXX %in% c(XXX, XXX)))  # Only include placebo and dimarta
  1. Using t.test() conduct a t-test comparing the change in histamine results between the adiclax (the standard of care) and dimarta. Did dimarta improve over the standard of care?

  2. Using glm() conduct a regression analysis predicting histamine_change as a function of 4 variables: treatment arm, histamine test results at the start of the trial, age, and quality of life at the start of the trial. Save the result as an object called histamine_change_glm. Once you do, apply the summary() function to the histamine_change_glm object to explore the results. Which variables reliably predict changes in test results?

  3. Repeat your previous regression analysis, but now predict change in quality of life. Do you get different results compared to your previous analysis?

  4. Now it’s time to see if the patient’s biomarkers were related to treatment success. For the dw biomarker, calculate the mean change in test results (histamine_change) separately for patients with different outcomes on the dw biomarker (hint: just group the data by dw and use summarise() to calculate the mean histamine_change).

  5. Do the same analysis for the other two biomarkers ms and np. Do either of these biomarkers seem to predict changes in test results?

  6. Using glm(), create a new regression object called histamine_change_bio_glm predicting histamine_change as a function of the 3 biomarkers. Explore the results with summary(). Do you find that any of these biomarkers predict changes in histamine?

  7. Did some drugs work better for patients with some biomarkers than others? For example, did patients who expressed the dw biomarker have better results when given dimarta compared to patients who do not express the dw biomarker? To answer this, start by calculating the descriptive statistics by calculating mean change in histamine histamine_change for all groups of dw and StudyArm (Hint: Just use group_by(dw, StudyArm) and summarise(histamine_change_mean = mean(histamine_change))).

  8. Once you’ve looked at the descriptive statistics, conduct a regression analysis predicting histamine_change based on the interaction between dw and StudyArm. Remember to calculate an interaction term in regression, use the * symbol in the formula. What do the results show?