R4CSR

Book club

This is a slide deck summary of R for Clinical Study Reporting and Submission

There are three main parts of the book:

  1. TLFs: a simulation of individual contributions

  2. Clinical trial projects: a simulation of project leadership

  3. eCTD submission: a simulation in package submission

In this summer book club we will get to just the first part, TLFs. Join us!

Get R help at UArizona

Data & Visualization Drop-in hours @ Main library

Data Science Institute workshops

CCT Data Science workshops

Preface

  • R folder structure recommended

  • CDISC pilotdata is used throughout

  • R packages needed

    • not all packages are used in every chapter
install.packages(c(
                    "dplyr"   # Manipulate data
                  , "emmeans" # Least-square mean estimation  
                  , "haven"  # import SAS data  
                  , "r2rtf" # Reporting in RTF format
                  , "survival" # kapplan meier curves
                  , "tidyr" # Manipulate data
                  , "table1" #Transfer data
                       ))

Living list of acronyms

Type Acronym Definition/Explanation
clinical CSR Clinical study report
clinical SDTM standard data tabulation model
clinical ADaM Analysis dataset model
clinical TLFs Tables, listings, and figures
clinical A&R Analysis and reporting
clinical/computational eCTD Electronic common technical document
clinical CDISC Clinical Data Interchange Standards Consortium
clinical ICH International conference on harmonization
clinical ICH E3 ICH guidelines on structure and content of clinical study reports
computational RTF Rich text format
clinical adae Analysis dataset for adverse events
clinical/computational ADSL Subject-level analysis dataset
clinical/computational ITT intention to treat (i.e. none of the patients are excluded and the patients are analyzed according to the randomization scheme)
statistical ANCOVA Analysis of covariance
statistical LOCF last observation carried forward
statistical K-M Kaplan Meier curve/estimate
clinical AEs adverse events

Chapter 1

An overview of clinical study reports and the {r2rtf} package.

Overview

  • A CSR contains all methods and results of a study in ~16 sections

    • TLFs are found in 10-12, 14, 16, +

  • often word document format

    • in this book we explore RTF

  • The {r2rtf} package allows for the creation and export of publication quality tables and figures in rich text format

    • Use the included r2rtf_adae dataset to explore functions in {r2rtf}

r2rtf

After formatting your data as desired by {r2rtf}, a general workflow is as follows:

head(tbl) %>%
  rtf_body() %>% # Step 1 Add table attributes
  rtf_encode() %>% # Step 2 Convert attributes to RTF encode
  write_rtf("tlf/intro-ae1.rtf") # Step 3 Write to a .rtf file

Each table attribute is added individually, then the attributes are converted to RTF, and finally, you can export an object in RTF.

Style with r2rtf

Use {r2rtf} to set specific design elements for your ouput:

head(tbl) %>%
  rtf_colheader(
    colheader = " | Treatment",
    col_rel_width = c(3, 6)
  ) %>%
  rtf_colheader(
    colheader = "Adverse Events | Placebo | Xanomeline High Dose | Xanomeline Low Dose",
    border_top = c("", "single", "single", "single"),
    col_rel_width = c(3, 2, 2, 2)
  ) %>%
  rtf_body(col_rel_width = c(3, 2, 2, 2)) %>%
  rtf_encode() %>%
  write_rtf("tlf/intro-ae7.rtf")

A single {r2rtf} command may include columns, borders, headers, width designations and many other elements.

Chapter 2

Section 10.1, Disposition of Participants

The disposition of participants table reports the numbers of participants who were randomized, and who entered and completed each phase of the study, and the reasons for all post-randomization discontinuations, grouped by treatment and by major reason (lost to follow-up, adverse event, poor compliance, etc.) are reported.

Disposition

Step 1: Count participants in the analysis population

Step 2: Calculate the number and percentage of participants who discontinued the study by treatment arm

Step 3: Calculate the numbers and percentages of participants who discontinued the study for different reasons, by treatment arm

Step 4: Calculate the number and percentage of participants who completed the study, by treatment arm

Step 5: Bind n_rand, n_disc, n_reason, and n_complete by row.

Step 6+: Write the final table to RTF

Disposition of participants RTF

tbl_disp %>%
  # Table title
  rtf_title("Disposition of Participants") %>%
  # First row of column header
  rtf_colheader(" | Placebo | Xanomeline Low Dose| Xanomeline High Dose",
    col_rel_width = c(3, rep(2, 3))
  ) %>%
  # Second row of column header
  rtf_colheader(" | n | (%) | n | (%) | n | (%)",
    col_rel_width = c(3, rep(c(0.7, 1.3), 3)),
    border_top = c("", rep("single", 6)),
    border_left = c("single", rep(c("single", ""), 3))
  ) %>%
  # Table body
  rtf_body(
    col_rel_width = c(3, rep(c(0.7, 1.3), 3)),
    text_justification = c("l", rep("c", 6)),
    border_left = c("single", rep(c("single", ""), 3))
  ) %>%
  # Encoding RTF syntax
  rtf_encode() %>%
  # Save to a file
  write_rtf("tlf/tbl_disp.rtf")

With our bound data we follow our workflow for {r2rtf}: add attributes, convert to RTF, write out

  • | separates every item, thus line 5 denotes 4 columns and line 9, 7 cols
  • line 6 could also read
    col_rel_width = c(3, 2, 2, 2)
  • and line 10 could also read
    col_rel_width = c(3, 0.7, 1.3, 0.7, 1.3, 0.7, 1.3)

Chapter 3

Section 11.1, Data Sets Analyzed

The summary of analysis sets table reports on participants included in each efficacy analysis

Writing functions

“You should consider writing a function whenever you’ve copied and pasted a block of code more than twice”

- R for Data Science


fmt_num <- function(x, digits, width = digits + 4) {
  formatC(x,
    digits = digits,
    format = "f",
    width = width
  )
}

Summary of analysis sets

With helper functions count_by and fmt_num we can more easily prepare the dataset for a summary of analysis sets table with the following steps:

Step 1: Bind the counts/percentages of the ITT population, the efficacy population, and the safety population by row using the count_by() function.
Step 2+: Format the output from Step 2 using r2rtf.

Chapter 4

Section 11.2 Demographic and other baseline characteristics

Creating a simple table to summarize critical demographic and baseline characteristics of the participants

R package {table1}

Efficiently summarizes this information and creates a HTML report


ana <- adsl %>%
  mutate(
    SEX = factor(SEX, c("F", "M"), c("Female", "Male")),
    RACE = toTitleCase(tolower(RACE))
  )

tbl <- table1(~ SEX + AGE + RACE | TRT01P, data = ana)
tbl

Transferring the data

Transferring the output into a dataframe that contains only ASCII characters, recommended by regulatory agencies

tbl_base <- tbl %>%
  as.data.frame() %>%
  as_tibble() %>%
  mutate(across(
    everything(),
    ~ str_replace_all(.x, intToUtf8(160), " ")
  ))


names(tbl_base) <- str_replace_all(names(tbl_base), intToUtf8(160), " ")
tbl_base

Final Formatting

Adjusting columns, headers, and indention.

colheader1 <- paste(names(tbl_base), collapse = "|")
colheader2 <- paste(tbl_base[1, ], collapse = "|")
rel_width <- c(2.5, rep(1, 4))

tbl_base[-1, ] %>%
  rtf_title(
    "Baseline Characteristics of Participants",
    "(All Participants Randomized)"
  ) %>%
  rtf_colheader(colheader1,
    col_rel_width = rel_width
  ) %>%
  rtf_colheader(colheader2,
    border_top = "",
    col_rel_width = rel_width
  ) %>%
  rtf_body(
    col_rel_width = rel_width,
    text_justification = c("l", rep("c", 4)),
    text_indent_first = -240,
    text_indent_left = 180
  ) %>%
  rtf_encode() %>%
  write_rtf("tlf/tlf_base.rtf")

Summary of steps

Step 1: Use package {table1}

Step 2: Transfer the output from Step 1 into an ASCII data frame

Step 3+: Define the format of the RTF table

Chapter 5

Section 11.4, Efficacy Results and Tabulations of Individual Participant

Summarizing primary and secondary endpoints

Efficacy Table

Combines two informative tables

  1. Summary of observed data

    • at baseline, week 24, and change from baseline

  2. Pairwise comparisons with placebo

Imputation

Missing data are inevitable

…why your data are missing can be highly informative

LOCF imputation is NOT a recommended imputation method

Read more about the prevention and treatment of missing data

ANCOVA

Analysis of covariance examines the realationship between an independent and dependent variable while controlling for a covariate

  • Specifically, compare variance around means of different groups

Use {emmeans} to calculate within and between group least square means

Efficacy tabulation steps

Step 1: Impute the missing values. In this example, we name the ana dataset after imputation as ana_locf.

Step 2: Calculate the mean and standard derivation of efficacy endpoint (i.e., gluc), and then format it into an RTF table.

Step 3: Calculate the pairwise comparison by ANCOVA model, and then format it into an RTF table.

Step 4: Combine the outputs from steps 4 and 5 by rows.

Step 5+: Format the output from Step 4 using r2rtf.

Chapter 6

Section 11.4, Efficacy Results and Tabulations of Individual Participant

The primary and secondary efficacy endpoints need to be summarized for each treatment group

Survival Analysis

aka time-to-event analysis


A survival model explores the relationship between an outcome variable and a censored estimate of time to study dropout (for any number of reasons).


fit <- survfit(Surv(timeToEvent, 1 - censoredTime) ~ treatmentGroup
               , data = clinicalData)

Efficacy figure steps

Step 1: Define the analysis-ready dataset

Step 2: Save figures into png files

Step 3+: Create RTF output

Chapter 7

Section 12.2 Adverse event (AEs) summary

Summarize adverse eventing, the number of patients in each treatment group in whom the event occurred, and the rate of occurrence.

Pivot wider

fuzzy monsters moving colorful rocks to demonstrate items in a dataframe moving between long and wide format

AE summary steps

Step 1: Summarize participants in population

Step 2: Summarize participants in population by AE criteria of interest

Step 3: Combine summaries

Step 4+: Format using r2rtf

Chapter 8

Section 12.2 Specific AEs

Report each adverse event, the number of patients in each treatment group in whom the event occurred, and the rate of occurrence.

page functions

The AE table introduces us to two advanced table features:

  • group content: AE can be summarized in multiple nested layers. (e.g., by system organ class (SOC, AESOC) and specific AE term (AEDECOD))

    • page_by()

  • pagenization: there are many AE terms that can not be covered in one page. Column headers and SOC information need to be repeated on every page.

    • pageby_row()

AE tabulation steps

Step 1: Count the number of participants by SOC and treatment arm

Step 2: Count the number of participants in each AE term by SOC, AE term, and treatment arm

Step 3: Count the number of participants in each arm

Step 4: Combine counts

Step 5+: Format the output by using r2rtf

Chapter 9

All TLFs are then assembled by

  1. Combining RTF source code in individual files into one large RTF file.

  2. Leveraging the Toggle Fields feature in Microsoft Word to embed RTF files using hyperlinks.

Combine source code


tlf_path <- c(
  "tlf/tbl_disp.rtf", # Disposition table
  "tlf/tlf_eff.rtf", # Efficacy table
  "tlf/tlf_km.rtf" # K-M plot
)

r2rtf::assemble_rtf(
  input = tlf_path,
  output = "tlf/rtf-combine.rtf"
)

Leverage Microsoft Word

  • Absolute paths
folder icons in a directory tree with numerous examples of absolute (from root) vs relative (to another folder) paths
  • Alt + F9 to “Toggle Fields”

Thank you!!