Background

Early in my career in the pharmaceutical industry, I quickly realized the substantial workload involved in late-stage clinical trials, particularly for global regulatory submissions spanning multiple studies. Efficient tools and workflows are essential for timely eCTD submissions after database lock.

Many trial statisticians use R for clinical study report (CSR) verification, often developing reusable functions and packages to improve efficiency. A key turning point for me was when Keaven asked me a simple yet insightful question: What is missing to make R viable for clinical study reports and regulatory submission? That same year, the R/Pharma conference was launched to explore open-source solutions.

Nan and I believed any sustainable solution must integrate seamlessly into existing, well-established workflows. With that principle in mind, we developed two essential R packages: r2rtf for reporting and pkglite for eCTD submission. We documented our approach in the R for Clinical Study Reports and Submission (R4CSR) book and contributed to the R Consortium R Submission Working Group’s first R submission pilot to the US FDA.

After joining Meta, I was encouraged by my colleagues to learn Python. With Nan’s parallel work on pkglite for Python, it was a natural step for me to explore Python-based solutions for clinical study reporting and submission—continuing our open-source efforts in a new ecosystem.

Example

Below is an example demonstrating how rtflite can be used to generate a baseline characteristics table in RTF format, similar to r2rtf:

from importlib.resources import files

import pandas as pd

import rtflite as rtf

data_path = files("rtflite.data").joinpath("baseline.csv")
df = pd.read_csv(data_path, na_filter=False)

header1 = pd.DataFrame([["", "Placebo", "Drug Low Dose", "Drug High Dose", "Total"]])
header2 = pd.DataFrame([["", "n", "(%)", "n", "(%)", "n", "(%)", "n", "(%)"]])

doc = rtf.RTFDocument(
    df=df,
    rtf_title=rtf.RTFTitle(
        text=["Demographic and Anthropometric Characteristics", "ITT Subjects"]
    ),
    rtf_column_header=[
        rtf.RTFColumnHeader(df=header1, col_rel_width=[3] + [2] * 4),
        rtf.RTFColumnHeader(
            df=header2,
            col_rel_width=[3] + [1.2, 0.8] * 4,
            border_top=[""] + ["single"] * 8,
            border_left=["single"] + ["single", ""] * 4,
        ),
    ],
    rtf_body=rtf.RTFBody(
        page_by=["var_label"],
        col_rel_width=[3] + [1.2, 0.8] * 4 + [3],
        text_justification=["l"] + ["c"] * 8 + ["l"],
        text_format=[""] * 9 + ["b"],
        border_left=["single"] + ["single", ""] * 4 + ["single"],
        border_top=[""] * 9 + ["single"],
        border_bottom=[""] * 9 + ["single"],
    ),
)

doc.write_rtf("output.rtf")

converter = rtf.LibreOfficeConverter()
converter.convert(
    input_files="output.rtf", output_dir=".", format="pdf", overwrite=True
)

Roadmap

Development of rtflite is structured into three phases:

• Phase 1: Support for single-page TLF generation (80% completed in v0.1.0).
• Phase 2: Implementation of pagination algorithms and features.
• Phase 3: Reproduction of all examples from the R4CSR book.

Beyond phase 3, we plan to showcase rtflite alongside pkglite for eCTD submissions in Python, replicating the success of the R submission pilots. We hope this work will contribute to the growing adoption of open-source solutions for regulatory-compliant analysis, reporting, and submission in drug and medical device development.

Acknowledgements

We extend our gratitude to Ross Farrugia, the pharmaverse council, and the pharmaverse community for their support in GitHub hosting and incorporating rtflite into the pharmaverse ecosystem.