Clinical Trial Data Analysis Using R: 13,748 Medical Trials Case Study with Code, Charts and Insights

Clinical Trial Data Analysis Using R: Complete SEO-Based Case Study

Clinical Trial Data Analysis Using R is a practical healthcare analytics project that teaches how to clean, explore, summarize and visualize real-world clinical trial data. This post uses a dataset of 13,748 medical trials and keeps the workflow suitable for beginners, students, healthcare researchers, data analysts, biostatistics learners and pharmaceutical data science portfolios.

This SEO-structured version preserves the complete original case study, including all tables, figures, links, code snippets, dataset values, statistical summaries and final findings. The structure has been improved for search engines by adding a clear focus keyword, metadata, stronger headings, an optimized introduction, source context, FAQ schema and article schema while keeping the original tutorial content intact.

A Real-World Case Study on 13,748 Medical Trials

---

Kaggle Clinical Trials Dataset

Clinical trial data analysis using R is one of the most practical applications of statistics, healthcare analytics, and data science. Instead of using toy datasets, this case study analyzes a real-world dataset containing more than 13,000 clinical trials from pharmaceutical and healthcare research.

In this guide, you will learn how to use R for:

• Clinical trial data cleaning
• Exploratory healthcare analytics
• Statistical summaries
• Trend analysis
• Medical data visualization
• Research reporting
• Real-world healthcare insights

Unlike generic tutorials, this article follows a complete analytical workflow used in actual healthcare and biostatistics environments.

Introduction to Clinical Trial Analytics

---

Clinical trials are the foundation of modern medical research. Pharmaceutical companies, hospitals, universities, and healthcare organizations conduct trials to test the following:

• Drug effectiveness
• Treatment safety
• Disease outcomes
• Medical interventions
• Patient recovery patterns

The global clinical research industry generates enormous datasets every year. These datasets contain structured information about the following: (TABLE 0)

Clinical Variable	Example
Trial Phase	Phases 1, 2, 3, 4
Enrollment Size	Number of Patients
Sponsor	Pfizer, Novartis
Disease Condition	Diabetes. Cancer
Status	Recurring, Completed
Timeline	Start year and month

Analyzing this information manually is impossible. This is where R becomes valuable.

Types of Clinical Trial Data

---

Understanding data types is essential before analysis.

Common Variables in Clinical Trials

Variable Type	Example
Continuous	Blood pressure
Binary	Survived / Did not survive
Categorical	Treatment group
Ordinal	Pain severity score
Time-to-Event	Survival duration

About the Dataset

---

This case study uses a public healthcare dataset from Kaggle containing more than 13,748 clinical trial records. (TABLE 2)

Dataset Columns

Column Name	Description
NCT	Unique Clinical
Sponsor	Organization conducting Study
Title	Trial Title
Summary	Trial Description
Start_Year	Year study Began
Start_Month	Month study began
Phase	Clinical Trial Phase
Enrollment	Number of Participants
Status	Current Trial Status
Condition	Disease or medical condition

Why Is This Dataset Valuable?

Unlike beginner datasets, this dataset reflects real-world healthcare complexity:

• Missing values
• Inconsistent categories
• Large text fields
• Multiple medical conditions
• Uneven enrollment distributions

These characteristics make it ideal for learning practical healthcare analytics.

Why Use R for Medical Data Analysis?

R is one of the most widely used tools in:

• Biostatistics
• Epidemiology
• Public health
• Clinical research
• Pharmaceutical analytics

Advantages of R in Healthcare

Feature	Benefit
Open Source	No licensing Cost
Statistical Power	Advanced Medical Analysis
Visualization	Publication-quality Charts
Reproducibility	Research Transparency
Community Support	Thousands of Packages
Automation	Faster Analysis Pipelines

Major healthcare organizations use R alongside SAS and Python. (TABLE 3)

Install R Programming Language

---

Go to the official CRAN website:

CRAN R Project

Installation Steps

1. Open the CRAN website
2. Click Download R for Windows
3. Click base
4. Click Download R for Windows
5. Run the installer
6. Keep default settings during installation

After installation: Open R and You should see the R Console window. (FIGURE 1)

Install RStudio

While R is the programming engine, RStudio provides a modern workspace for coding and analysis.

RStudio makes it easier to:

• Write scripts
• Import datasets
• View tables
• Create charts
• Manage projects
• Export reports

Download from the official website:

Posit RStudio Download

Installation Steps

Open RStudio after installation

Download RStudio Desktop

Run the installer

Keep default installation settings

When RStudio opens, you will see four main panels. (FIGURE 2)

Panel	Purpose
Script Editor	Write R Code
Console	Execute Commands
Environment	View Variables
Files/Plots	View Charts and Files

Create an RStudio project

Open RStudio File → New Project
Existing Directory/Create New Directory
Browse to:
F:\CLINICAL_DATA  (FOLDER FOR THE PROJECT OR CREARTE NEW)
Create Project
Now all files, scripts, and outputs will remain organized in one workspace. (FIGURE 3)

Create Your First R Script

---

Inside RStudio:

Click: File>New File> R Script

Save the script as: clinical_trial_analysis.R (FIGURE 4)

This script will contain all analysis code used in this case study.

A clean project structure improves workflow efficiency.

F:\CLINICAL_DATA/
│
├── AERO-BirdsEye-Data.csv
├── clinical_trial_analysis.R
├── plots/
├── outputs/
└── reports/

Common Beginner Mistakes During Installation

MISTAKE	SOLUTION
Installing RStudio before R	Install R first
Spaces in file paths causing issues	Use quotes around paths
Forgetting to save scripts	Save frequently
Mixing datasets and scripts	Use folders

Install Core Packages

---

In clinical_trial_analysis.R copy and paste the following code to install packages.

Task	Example Package
Data cleaning	dplyr
Visualization	ggplot2
Data Importing	readr
Healthcare analytics	survival
Interactive charts	plotly

install.packages(c( “tidyverse”, “ggplot2”, “dplyr”, “readr”, “janitor”, “stringr”, “lubridate”, “plotly” ))

After installation, packages must be loaded into the session. Add this code below the installation section. (FIGURE 5)

library(tidyverse) library(ggplot2) library(dplyr) library(readr) library(janitor) library(stringr) library(lubridate) library(plotly)

Understanding Each Package

tidyverse

Tidyverse is a collection of modern R packages for data science.

It simplifies:

• Data cleaning
• Filtering
• Transformation
• Visualization

Why It Matters

Most professional R workflows now rely on the tidyverse ecosystem.

ggplot2

ggplot2 is one of the most powerful visualization libraries in data science.

We will use it to create:

• Trial phase charts
• Enrollment distributions
• Sponsor visualizations
• Clinical research trend graphs

dplyr

dplyr is used for:

• Filtering data
• Grouping observations
• Creating summaries
• Data transformation

Example tasks:

• Counting trial phases
• Finding top sponsors
• Grouping disease conditions

readr

readr helps import datasets quickly and efficiently.

We will use it to load the clinical trial CSV file.

janitor

janitor helps clean messy column names and improve dataset consistency.

Healthcare datasets often contain inconsistent formatting, making this package extremely useful.

stringr

stringr simplifies text cleaning and manipulation.

We will use it for:

• Cleaning sponsor names
• Standardizing condition labels
• Formatting text variables

lubridate

lubridate simplifies working with dates and time variables.

Useful for:

• Start year analysis
• Timeline visualization
• Trend analysis

plotly

plotly allows interactive charts.

This can improve blog engagement if you later embed visualizations online.

Importing Clinical Trial Data

---

Clinical datasets are commonly stored in CSV, Excel, or SAS formats.

Import CSV File

trial_data <- read.csv(“AERO-BirdsEye-Data.csv”)

Import Excel File

library(readxl)
trial_data <- read_excel(“AERO-BirdsEye-Data.csv”)

View Dataset Structure

head(trial_data)

It displays first rows. (FIGURE 06)

Check Dataset Dimensions

---

Understanding dataset size is essential before analysis.

dim(trial_data)

Expected output:

13748 11

Meaning:

Value	Meaning
13,748	Rows
11	Columns

View Column Names

Column names help identify variables available for analysis.

colnames(trial_data)

Expected variables:

Column Name	Purpose
NCT	Trial identifier
Sponsor	Organization
Title	Study title
Summary	Study description
Start_Year	Trial year
Start_Month	Trial month
Phase	Trial phase
Enrollment	Number of patients
Status	Current state
Condition	Medical condition

---

Understand Dataset Structure

Use str() to inspect variable types.

str(trial_data)

This shows:

• Numeric variables
• Character variables
• Missing values
• Data structure

str(trial_data) ‘data.frame’: 13748 obs. of 11 variables: $ index : int 0 1 2 3 4 5 6 7 8 9 … $ NCT : chr “NCT00003305” “NCT00003821” “NCT00004025” “NCT00005645” … $ Sponsor : chr “Sanofi” “Sanofi” “Sanofi” “Sanofi” … $ Title : chr “A Phase II Trial of Aminopterin in Adults and Children With Refractory Acute Leukemia Grant Application Title: “| __truncated__ “Phase II Trial of Aminopterin in Patients With Persistent or Recurrent Endometrial Carcinoma” “Phase I/II Trial of the Safety, Immunogenicity, and Efficacy of Autologous Dendritic Cells Transduced With Aden”| __truncated__ “Phase II Trial of ILX295501 Administered Orally Once Weekly x 3 Repeated Every 6 Weeks in Patients With Stage I”| __truncated__ … $ Summary : chr “RATIONALE: Drugs used in chemotherapy use different ways to stop cancer cells from dividing so they stop growin”| __truncated__ “RATIONALE: Drugs used in chemotherapy use different ways to stop tumor cells from dividing so they stop growing”| __truncated__ “RATIONALE: Vaccines made from a person’s white blood cells combined with melanoma antigens may make the body bu”| __truncated__ “RATIONALE: Drugs used in chemotherapy use different ways to stop tumor cells from dividing so they stop growing”| __truncated__ … $ Start_Year : int 1997 1998 1999 1999 2000 2000 2000 2000 2001 2001 … $ Start_Month: int 7 1 3 5 10 11 12 7 4 5 … $ Phase : chr “Phase 2” “Phase 2” “Phase 1/Phase 2” “Phase 2” … $ Enrollment : int 75 0 36 0 0 0 0 0 3222 8 … $ Status : chr “Completed” “Withdrawn” “Unknown status” “Withdrawn” … $ Condition : chr “Leukemia” “Endometrial Neoplasms” “Melanoma” “Ovarian Neoplasms” … >

Why Structure Inspection Is Important

Clinical datasets often import incorrectly.

Common Problems

Problem	Example
Numbers imported as text	Enrollment column
Missing values misread	Empty cells
Dates stored incorrectly	Start year/month
Mixed categories	Status labels

Inspecting structure early prevents future statistical errors.

Generate a Statistical Summary

Now generate summary statistics.

summary(trial_data)

This provides:

• Minimum values
• Maximum values
• Mean
• Median
• Missing values

Key Findings from the Statistical Summary

---

The statistical summary gives several useful insights before cleaning the dataset.

The Dataset Contains 13,748 Clinical Trial Records

The summary output confirms that most columns contain 13,748 entries, which means the dataset is large enough for meaningful exploratory data analysis.

This is important because a larger dataset allows us to study patterns in:

• Trial phases
• Enrollment size
• Medical conditions
• Sponsor activity
• Trial status
• Research trends over time

For a real-world clinical trial case study, 13,748 records provide a strong foundation for analysis.

Start Year Ranges from 1984 to 2020

The Start_Year column shows the following range:

Statistic	Value
Minimum Year	1984
1st Quartile	2006
Median Year	2009
Mean Year	2009
3rd Quartile	2013
Maximum Year	2020

This tells us that the dataset covers clinical trials from 1984 to 2020.

However, most trials appear to be concentrated after 2006 because the first quartile is 2006. The median year is 2009, which means half of the trials started before 2009 and half started after 2009.

This is useful for our later time-trend analysis, where we can examine how clinical trial activity changed across different years.

Start Month Looks Properly Distributed

The Start_Month column ranges from 1 to 12, which means the month variable appears to be correctly formatted.

Statistic	Value
Minimum Month	1
Median Month	7
Mean Month	6.691
Maximum Month	12

This confirms that the dataset uses standard month numbers, where:

Month Number	Meaning
1	January
6	June
12	December

The median start month is 7, meaning July is around the center of the distribution. This column can later be used to analyze whether clinical trials are more commonly started in certain months.

---

Enrollment Has a Very Wide Range

The Enrollment column is one of the most important variables in this dataset because it tells us how many participants were included in each clinical trial.

The summary shows:

Statistic	Enrollment
Minimum	0
1st Quartile	40
Median	124
Mean	440.8
3rd Quartile	365
Maximum	84,496

This is one of the most important findings so far.

The median enrollment is 124, but the mean enrollment is 440.8. This difference shows that the enrollment data is likely right-skewed.

In simple terms, most trials have a relatively small number of participants, but a few very large trials increase the average.

This is common in clinical research. Some early-phase studies may include only a small number of patients, while large Phase 3 or Phase 4 trials can include thousands of participants.

The maximum enrollment value of 84,496 is extremely high and should be inspected later as a possible outlier or large-scale observational study.

---

Phase Column Has 8 Unique Categories

The Phase column has:

Metric	Value
Total Records	13,748
Unique Values	8
Blank Values	253

The presence of 8 unique phase categories means that this dataset may include values such as:

• Phase 1
• Phase 2
• Phase 3
• Phase 4
• Phase 1/Phase 2
• Phase 2/Phase 3
• Not Applicable
• Blank or missing values

There are also 253 blank phase values, which means this column will need cleaning before analysis.

This matters because clinical trial phase is a key variable for understanding the purpose and maturity of each study.

Status Column Has 9 Unique Categories

The Status column contains 9 unique values.

This means the dataset likely includes trial statuses such as:

• Completed
• Recruiting
• Terminated
• Withdrawn
• Active
• Suspended
• Unknown

This is useful because status analysis can help us understand how many trials were successfully completed and how many were stopped or withdrawn.

Later in this case study, we will create a frequency table and bar chart to show the distribution of clinical trial statuses.

---

Condition Column Has 867 Unique Medical Conditions

The Condition column contains 867 unique values.

This is a strong signal that the dataset covers a wide range of diseases and healthcare research areas.

The summary shows:

Metric	Value
Total Records	13,748
Unique Conditions	867
Minimum Character Length	4
Maximum Character Length	58

This column will be especially useful for finding the most commonly studied medical conditions.

Later, we can use this variable to answer questions such as:

• Which diseases appear most often in this clinical trial dataset?
• Are cancer-related trials common?
• How many trials focus on diabetes, asthma, hypertension, or cardiovascular disease?
• Which medical conditions attract the most research attention?

This section will give the post strong SEO value because people often search for real-world healthcare analytics examples using R.

Text Columns Need Further Inspection

The Title and Summary columns are text-based variables.

The summary shows that:

Column	Maximum Character Length
Title	591
Summary	4951

This means some clinical trial summaries are very detailed. These text columns may be useful for advanced analysis later, such as:

• Keyword extraction
• Text mining
• Natural language processing
• Identifying treatment areas
• Classifying trials by topic

For this beginner-friendly case study, we will mainly use structured columns such as phase, status, enrollment, year, sponsor, and condition.

What This Summary Tells Us Before Data Cleaning

The statistical summary gives us a clear understanding of the dataset before deeper analysis.

Main Observations

Observation	Meaning
Dataset has 13,748 records	Good size for real-world analysis
Trials range from 1984 to 2020	Useful for time-trend analysis
Enrollment is highly skewed	Outlier analysis is needed
Phase has blank values	Data cleaning is required
Condition has 867 unique values	Strong disease-level analysis possible
Status has 9 categories	Trial completion analysis is possible
Text columns are large	Future NLP analysis is possible

Why This Step Matters in Clinical Trial Data Analysis Using R

Before creating charts or running statistical models, we need to understand the raw dataset.

This step helps prevent common mistakes such as:

• Analyzing missing values incorrectly
• Treating text as numeric data
• Ignoring outliers
• Misinterpreting categorical variables
• Creating misleading charts

In healthcare analytics, small data quality issues can lead to incorrect conclusions. That is why every clinical trial data analysis project should begin with a proper summary and interpretation.

Cleaning the Clinical Trial Dataset in R

---

After checking the statistical summary, the next step is data cleaning. In real-world clinical trial data analysis using R, raw datasets are rarely ready for analysis. They often contain blank values, inconsistent column names, repeated records, irregular text formatting, and unusual numeric values.

Before creating charts or drawing conclusions, we need to prepare a clean version of the dataset.

In this step, we will:

• Clean column names
• Remove duplicate records
• Convert blank values into missing values
• Standardize text columns
• Check the structure of the cleaned dataset

This process helps make the analysis more reliable and easier to reproduce.

Why Data Cleaning Matters in Clinical Trial Analysis

Clinical trial datasets contain information about medical studies, sponsors, enrollment, conditions, and trial phases. If the dataset is not cleaned properly, the final results may be misleading.

For example:

Data Issue	Why It Matters
Blank phase values	Phase analysis becomes inaccurate
Duplicate records	Trial counts become inflated
Inconsistent sponsor names	Sponsor ranking becomes unreliable
Enrollment outliers	Average enrollment becomes misleading
Text formatting issues	Grouping and filtering become harder

In healthcare analytics, clean data is especially important because research decisions may depend on the results.

Create a Backup Copy of the Original Dataset

Before cleaning, it is best practice to keep the original dataset unchanged.

Run this command:

trial_raw <- trial_data

This creates a backup called trial_raw.

Now, if we make any mistake during cleaning, the original imported dataset is still available.

Clean Column Names

Dataset may have column names with spaces, capital letters, or special characters. These can make R code harder to write.

We will use the clean_names() function from the janitor package.

trial_clean <- trial_data %>%
  clean_names()

This will convert column names into a cleaner format.

For example:

Original Column	Cleaned Column
Start_Year	start_year
Start_Month	start_month
Enrollment	enrollment
Condition	condition
Sponsor	sponsor

Clean column names are easier to use in R code and reduce errors during analysis.

Check the New Column Names

Run this command:

names(trial_clean)

You should now see lowercase column names such as:

"[1] "index"       "nct"         "sponsor"     "title"       "summary"     "start_year"  "start_month" "phase"      
 [9] "enrollment"  "status"      "condition"  

This confirms that the column names are ready for analysis.

Convert Blank Values into NA

In many CSV files, missing values are stored as blank cells instead of proper NA values. R may not always detect these automatically.

To fix this, run:

trial_clean <- trial_clean %>%
  mutate(across(where(is.character), ~na_if(.x, "")))

This command checks all text columns and converts blank values into NA.

That makes missing value analysis more accurate in the next step.

Remove Duplicate Records

Duplicate records can inflate trial counts and distort analysis.

Run:

trial_clean <- trial_clean %>%
  distinct()

This keeps only unique rows in the dataset.

Standardize text columns.

Text values should be consistent before grouping and counting.

Run:

trial_clean <- trial_clean %>%
  mutate(
    sponsor = str_squish(str_to_title(sponsor)),
    phase = str_squish(phase),
    status = str_squish(status),
    condition = str_squish(condition)
  )

This code does four useful things:

Cleaning Function	Purpose
str_squish()	Removes extra spaces
str_to_title()	Standardizes sponsor capitalization
mutate()	Updates selected columns
condition = str_squish(condition)	Cleans condition text

This is especially useful for sponsor analysis. For example, it helps avoid treating similar sponsor names as different categories because of spacing or capitalization.

Comparing Raw Data and Cleaned Data Before Analysis

---

Before moving into charts and deeper clinical trial data analysis using R, it is important to understand the difference between the raw dataset and the cleaned dataset.

In the previous step, we cleaned the imported dataset and created a new object called

trial_clean

From this point forward, we should use trial_clean for all analysis instead of the original trial_data.

The raw dataset is useful for initial inspection, but the cleaned dataset is better for final analysis, visualization, and interpretation.

What Is Raw Data?

Raw data is the dataset exactly as it was imported from the CSV file.

It has not yet been cleaned, standardized, or prepared for analysis.

In this case study, the raw clinical trial dataset contained:

Feature	Raw Data Observation
Total records	13,748
Start year range	1984 to 2020
Start month range	1 to 12
Enrollment range	0 to 84,496
Phase categories	8 unique values
Blank phase values	253
Status categories	9 unique values
Condition categories	867 unique values

This tells us that the dataset is large and useful for healthcare analytics, but it also contains data quality issues that should be handled before analysis.

What Is Cleaned Data?

Cleaned data is the processed version of the raw dataset.

In this project, cleaning included:

• Standardizing column names
• Converting blank values into proper NA values
• Removing duplicate rows
• Cleaning extra spaces from text columns
• Creating analysis-ready variables

The cleaned dataset is stored as:

trial_clean

This is the version we will use for the rest of the case study.

Raw Data vs Cleaned Data

Feature	Raw Data	Cleaned Data
Object name in R	trial_data	trial_clean
Source	Directly imported CSV file	Processed version of raw data
Column names	Mixed capitalization	Standardized lowercase names
Blank values	May appear as empty cells	Converted into NA
Duplicate rows	May still be present	Checked and removed
Text formatting	May contain extra spaces	Cleaned using string functions
Best use	Initial inspection	Final analysis and visualization

The raw dataset helps us understand the original file.
The cleaned dataset helps us produce reliable analysis.

Why We Should Not Analyze Raw Data Directly

If we analyze the raw dataset without cleaning it, the results may be inaccurate.

Raw Data Problem	Possible Effect
Blank phase values	Phase counts become unclear
Duplicate rows	Trial totals may become inflated
Extra spaces in sponsor names	Same sponsor may appear as different sponsors
Enrollment value of 0	Mean enrollment may become misleading
Inconsistent text formatting	Grouping and filtering become inaccurate

In clinical trial data analysis using R, cleaning is not optional. It is a necessary step before creating final tables, charts, or conclusions.

Interpreting the Cleaned Clinical Trial Dataset Summary

After cleaning the dataset, we again used the summary() function to check whether the data was ready for analysis.

summary(trial_clean)

This step is important because data cleaning should not be done blindly. After cleaning, we need to verify what changed, what remained the same, and whether the dataset is suitable for clinical trial data analysis using R.

The cleaned dataset still contains 13,748 records, which means no complete rows were removed during the cleaning process. However, the summary shows important improvements in how missing and blank values are represented.

Cleaned Dataset Overview

The cleaned dataset includes the same core clinical trial variables:

Variable	Description
index	Row index from the original dataset
nct	Clinical trial identification number
sponsor	Organization responsible for the trial
title	Clinical trial title
summary	Description of the trial
start_year	Year the clinical trial started
start_month	Month the clinical trial started
phase	Clinical trial phase
enrollment	Number of participants
status	Trial status
condition	Medical condition studied

The biggest visible change is that column names are now cleaner and easier to use in R. For example, Start_Year became start_year, and Start_Month became start_month.

This makes the code easier to write and reduces the chance of errors.

Raw Data vs Cleaned Data: What Changed?

Feature	Raw Dataset: trial_data	Cleaned Dataset: trial_clean	What Changed
Total rows	13,748	13,748	Same number of records
Column names	Mixed capitalization	Lowercase clean names	Improved for coding
Title blank values	144 blanks	144 NA values	Blanks converted to missing values
Phase blank values	253 blanks	263 NA values	Missing phases now detected properly
Summary blanks	0 blanks	0 blanks	No major change
Sponsor unique values	10	10	Same sponsor groups
Status unique values	9	9	Same status groups
Condition unique values	867	867	Same disease categories
Enrollment range	0 to 84,496	0 to 84,496	Same numeric range

The most important improvement is that blank values are now treated as real missing values using NA.

This is important because R handles NA values properly in summaries, charts, filters, and statistical analysis.

Start Year Remains Valid

The start_year column still ranges from 1984 to 2020.

Statistic	Value
Minimum	1984
1st Quartile	2006
Median	2009
Mean	2009
3rd Quartile	2013
Maximum	2020

This confirms that cleaning did not change the historical coverage of the dataset.

The dataset still gives us a useful timeline of clinical trial activity over several decades. Later, we will use this column to create a trend chart showing how clinical trial activity changed by year.

Enrollment Still Shows Strong Skewness

The enrollment column remains unchanged after cleaning.

Statistic	Value
Minimum	0
1st Quartile	40
Median	124
Mean	440.8
3rd Quartile	365
Maximum	84,496

This tells us that the enrollment distribution is highly uneven.

The median enrollment is 124, but the mean is 440.8. This difference shows that a small number of very large clinical studies are pulling the average upward.

This is common in real-world clinical trial datasets because some studies enroll only a few dozen participants, while others include thousands.

Phase Has Missing Values After Cleaning

The cleaned summary shows that the phase column has 263 missing values.

Metric	Value
Unique phase values	7
Missing phase values	263

This means the dataset contains multiple clinical trial phase categories, but some records do not include phase information.

This can happen when studies are observational, device-related, non-drug studies, or not assigned to a traditional clinical phase.

For this case study, the best approach is not to delete these records. Instead, we will label missing phase values as Not Available

Checking Missing Values in the Cleaned Dataset

---

After cleaning the dataset and converting blank values into proper NA values, the next step is to check missing values by column.

Missing value analysis is important in clinical trial data analysis using R because missing data can affect charts, frequency tables, statistical summaries, and final conclusions.

We used:

colSums(is.na(trial_clean))

The output was:

Column	Missing Values
index	0
nct	0
sponsor	0
title	144
summary	0
start_year	0
start_month	0
phase	263
enrollment	0
status	0
condition	0

This is a useful result because most important analytical columns do not have missing values.

What These Missing Values Mean

The cleaned dataset contains missing values in only two columns:

1. title
2. phase

All other columns are complete.

This means the dataset is in good condition for analysis because the main variables needed for clinical trial trend analysis are available.

Title Has 144 Missing Values

The title column has 144 missing values.

Column	Missing Values	Impact
title	144	Low impact for this analysis

This is not a major issue for our current case study because we are not using the title column for statistical analysis.

The title column is mainly useful for:

• Text mining
• Keyword extraction
• Searching individual studies
• Natural language processing
• Identifying study themes

Since our current analysis focuses on structured variables such as phase, enrollment, status, sponsor, year, and condition, missing titles will not affect most of our results.

Phase Has 263 Missing Values

The phase column has 263 missing values.

Column	Missing Values	Impact
phase	263	Important for phase analysis

This is more important than missing titles because clinical trial phase is one of the key variables in this case study.

If phase values are missing, the phase distribution chart may be incomplete unless we handle those missing values properly.

Why We Should Not Use na.omit() Here

A beginner might remove all missing values using:

na.omit(trial_clean)

But that is not the best choice here.

If we remove every row with a missing title or missing phase, we may lose useful clinical trial records that still contain valid sponsor, year, enrollment, status, and condition information.

In healthcare analytics, removing data without a clear reason can reduce accuracy and weaken the analysis.

A better approach is to handle missing phase values transparently.

Creating a Clean Clinical Trial Phase Variable

---

Because the phase column had missing values, we created a new variable called phase_clean.

This column keeps existing phase values but replaces missing values with:

Not Available

We used:

trial_clean <- trial_clean %>%
  mutate(
    phase_clean = ifelse(is.na(phase), "Not Available", phase)
  )

Then we checked the distribution:

table(trial_clean$phase_clean)

The output was:

Clinical Trial Phase	Number of Trials
Early Phase 1	10
Not Available	263
Phase 1	2,516
Phase 1/Phase 2	322
Phase 2	3,596
Phase 2/Phase 3	139
Phase 3	4,887
Phase 4	2,015

---

Why We Created phase_clean

The original phase column had missing values. If we used it directly in charts, R might silently exclude missing values or show them unclearly.

By creating phase_clean, we make the analysis easier to understand.

Original Issue	Cleaning Solution
Missing phase values	Replaced with Not Available
Harder chart interpretation	Clear category for missing data
Possible row loss	All records preserved
Less transparent analysis	Missing data shown openly

From this point forward, phase-based analysis should use:

phase_clean

instead of:

phase

Creating a Clean Phase Distribution Table in R

---

After creating the phase_clean variable, we converted the phase counts into a structured table.

We used:

phase_table <- trial_clean %>%
  count(phase_clean) %>%
  arrange(desc(n))

phase_table

The result was:

Rank	Clinical Trial Phase	Number of Trials
1	Phase 3	4,887
2	Phase 2	3,596
3	Phase 1	2,516
4	Phase 4	2,015
5	Phase 1/Phase 2	322
6	Not Available	263
7	Phase 2/Phase 3	139
8	Early Phase 1	10

This table gives a clearer view of clinical trial phase distribution.

---

Phase 3 Has the Highest Number of Trials

The most common category is Phase 3, with 4,887 trials.

Phase 3 trials are usually larger and more advanced studies designed to confirm whether a treatment works in a broader patient population.

A high number of Phase 3 trials suggests that the dataset includes many late-stage clinical studies.

Phase 2 Is the Second-Largest Category

The second-largest category is Phase 2, with 3,596 trials.

Together, Phase 2 and Phase 3 make up a major portion of the dataset:

Combined Category	Number of Trials
Phase 2 + Phase 3	8,483

This shows that most studies in this dataset are concentrated in mid-stage and late-stage clinical research.

Visualizing the Clinical Trial Phase Distribution

---

After creating the phase distribution table, we visualized the results using a bar chart.

ggplot(phase_table, aes(x = reorder(phase_clean, n), y = n)) +
  geom_col() +
  coord_flip() +
  labs(
    title = "Clinical Trial Phase Distribution",
    x = "Clinical Trial Phase",
    y = "Number of Trials"
  ) +
  theme_minimal()

This chart gives a clearer visual understanding of the distribution of clinical trial phases in the dataset.

What the Bar Chart Shows

The chart confirms the same pattern we observed in the table.

Phase 3 has the longest bar, followed by Phase 2, Phase 1, and Phase 4.

This shows that the dataset is not evenly distributed across phases. Instead, it is concentrated mainly in:

• Phase 3
• Phase 2
• Phase 1
• Phase 4

The chart also shows that missing phase information is limited because the Not Available category is relatively small.

Calculating the Percentage Share of Each Clinical Trial Phase

---

Counts are useful, but percentages make the results easier to compare.

We used:

phase_table <- phase_table %>%
  mutate(percentage = round((n / sum(n)) * 100, 2))

phase_table

The result was:

Rank	Clinical Trial Phase	Number of Trials	Percentage
1	Phase 3	4,887	35.55%
2	Phase 2	3,596	26.16%
3	Phase 1	2,516	18.30%
4	Phase 4	2,015	14.66%
5	Phase 1/Phase 2	322	2.34%
6	Not Available	263	1.91%
7	Phase 2/Phase 3	139	1.01%
8	Early Phase 1	10	0.07%

Phase 2 and Phase 3 Make Up Most of the Dataset

The two largest categories are:

Phase	Percentage
Phase 3	35.55%
Phase 2	26.16%
Combined	61.71%

Together, Phase 2 and Phase 3 represent 61.71% of the dataset.

This means nearly two-thirds of all records are concentrated in mid-stage and late-stage clinical research.

Interpreting the Percentage Distribution of Clinical Trial Phases

---

We also created a percentage-based bar chart.

ggplot(phase_table, aes(x = reorder(phase_clean, percentage), y = percentage)) +
  geom_col() +
  coord_flip() +
  labs(
    title = "Percentage Distribution of Clinical Trial Phases",
    x = "Clinical Trial Phase",
    y = "Percentage of Trials"
  ) +
  theme_minimal()

The chart makes one conclusion very clear: this clinical trial dataset is heavily concentrated in Phase 2 and Phase 3 studies.

Key Insights from Phase Analysis

Finding	Interpretation
Phase 3 is 35.55%	Largest clinical trial category
Phase 2 is 26.16%	Strong focus on efficacy testing
Phase 2 + Phase 3 = 61.71%	Dataset is dominated by mid-to-late stage trials
Phase 1 is 18.30%	Early-stage safety trials are well represented
Phase 4 is 14.66%	Post-market research is also meaningful
Not Available is 1.91%	Missing phase data is limited
Early Phase 1 is 0.07%	Very early studies are rare

Analyzing Clinical Trial Status Distribution in R

---

After analyzing clinical trial phases, the next step is to examine the status of each clinical trial.

The status column tells us whether a clinical trial was completed, terminated, withdrawn, suspended, recruiting, or still active.

We used:

status_table <- trial_clean %>%
  count(status) %>%
  arrange(desc(n))

status_table

The result was:

Rank	Clinical Trial Status	Number of Trials
1	Completed	10,568
2	Terminated	1,285
3	Recruiting	800
4	Active, not recruiting	646
5	Withdrawn	291
6	Not yet recruiting	108
7	Unknown status	19
8	Suspended	16
9	Enrolling by invitation	15

---

Most Trials Were Completed

The largest category is Completed, with 10,568 trials.

This shows that the majority of clinical trials in this dataset reached completion.

Completed studies are especially useful for historical analysis because they are more likely to have full enrollment information and final trial records.

Terminated Trials Are the Second-Largest Category

The second-largest category is Terminated, with 1,285 trials.

A terminated trial is a study that started but ended earlier than planned.

This may happen for several reasons, such as:

• Safety concerns
• Low recruitment
• Lack of effectiveness
• Funding issues
• Administrative decisions
• Sponsor strategy changes

This is an important insight because terminated trials can reveal challenges in clinical research.

Calculating the Percentage Distribution of Clinical Trial Status

We calculated the percentage share of each status.

status_table <- status_table %>%
  mutate(percentage = round((n / sum(n)) * 100, 2))

status_table

The result was:

Rank	Clinical Trial Status	Number of Trials	Percentage
1	Completed	10,568	76.87%
2	Terminated	1,285	9.35%
3	Recruiting	800	5.82%
4	Active, not recruiting	646	4.70%
5	Withdrawn	291	2.12%
6	Not yet recruiting	108	0.79%
7	Unknown status	19	0.14%
8	Suspended	16	0.12%
9	Enrolling by invitation	15	0.11%

What the Status Percentage Table Shows

The most important result is:

76.87% of trials were completed.

This means more than three out of every four records in the dataset belong to completed clinical trials.

Terminated trials represent 9.35%, which means nearly one in ten trials ended earlier than planned.

Recruiting and active trials together represent 10.52%, showing that some studies were still ongoing at the time the dataset was collected.

Visualizing the Percentage Distribution of Clinical Trial Status

We created a bar chart for clinical trial status.

ggplot(status_table, aes(x = reorder(status, percentage), y = percentage)) +
  geom_col() +
  coord_flip() +
  labs(
    title = "Percentage Distribution of Clinical Trial Status",
    x = "Clinical Trial Status",
    y = "Percentage of Trials"
  ) +
  theme_minimal()

The chart shows that Completed trials dominate the dataset, while all other status categories are much smaller.

Main Insights from Trial Status Analysis

Finding	Interpretation
Completed trials = 76.87%	Most studies reached completion
Terminated trials = 9.35%	A meaningful number ended early
Recruiting + active trials = 10.52%	Some studies were ongoing
Withdrawn trials = 2.12%	A small number stopped before enrollment
Unknown status = 0.14%	Status data is mostly complete
Suspended trials = 0.12%	Temporary pauses are rare

Analyzing Clinical Trial Enrollment Summary in R

After analyzing trial phases and trial status, the next step is to examine the enrollment variable.

Enrollment tells us how many participants were included in each clinical trial.

We used:

summary(trial_clean$enrollment)

Then we created a structured summary:

enrollment_summary <- trial_clean %>%
  summarise(
    total_trials = n(),
    min_enrollment = min(enrollment, na.rm = TRUE),
    q1_enrollment = quantile(enrollment, 0.25, na.rm = TRUE),
    median_enrollment = median(enrollment, na.rm = TRUE),
    mean_enrollment = round(mean(enrollment, na.rm = TRUE), 2),
    q3_enrollment = quantile(enrollment, 0.75, na.rm = TRUE),
    max_enrollment = max(enrollment, na.rm = TRUE)
  )

enrollment_summary

The result was:

Total Trials	Minimum Enrollment	1st Quartile	Median	Mean	3rd Quartile	Maximum Enrollment
13,748	0	40	124	440.78	365	84,496

What the Enrollment Summary Shows

The enrollment summary shows that clinical trial size varies dramatically.

The median enrollment is 124, while the mean enrollment is 440.78.

This difference means enrollment is right-skewed. Most trials are relatively small, but a few very large studies increase the average.

The maximum enrollment value is 84,496, which is much larger than the median.

Why Median Is Better Than Mean Here

In skewed healthcare datasets, the median often gives a better picture of the typical case.

If we only reported the mean enrollment of 440.78, readers might assume most trials are around 441 participants.

But the median of 124 tells a different story.

It shows that a more typical clinical trial in this dataset has around 124 participants, while the mean is inflated by very large studies.

Visualizing the Distribution of Clinical Trial Enrollment

---

We created a histogram to understand how enrollment is distributed.

ggplot(trial_clean, aes(x = enrollment)) +
  geom_histogram(bins = 50) +
  labs(
    title = "Distribution of Clinical Trial Enrollment",
    x = "Number of Participants Enrolled",
    y = "Number of Trials"
  ) +
  theme_minimal()

The histogram shows that most trials are concentrated near the lower end of the enrollment range, while a few very large studies stretch the x-axis.

This confirms that clinical trial enrollment is highly right-skewed.

Interpreting the Log-Scale Distribution of Clinical Trial Enrollment

---

Because the enrollment distribution was extremely skewed, we created a log-scale histogram.

ggplot(trial_clean, aes(x = enrollment + 1)) +
  geom_histogram(bins = 50) +
  scale_x_log10() +
  labs(
    title = "Log-Scale Distribution of Clinical Trial Enrollment",
    x = "Number of Participants Enrolled (Log Scale)",
    y = "Number of Trials"
  ) +
  theme_minimal()

We used:

enrollment + 1

because log-scale charts cannot handle zero values.

Why We Used a Log Scale

A normal histogram was useful, but it was difficult to interpret because enrollment ranged from 0 to 84,496 participants.

A log scale helps by compressing very large values while spreading out smaller values.

Normal Scale	Log Scale
Shows the full raw range	Makes skewed data easier to read
Outliers dominate the chart	Outliers are compressed
Smaller trials are hard to see	Small and medium trials become visible
Useful first view	Better for interpretation

The log-scale chart shows that most trials are small to medium-sized, while very large studies are rare.

Identifying the Largest Clinical Trials by Enrollment

---

To understand which studies created the long right tail, we identified the top 10 trials by enrollment.

top_enrollment_trials <- trial_clean %>%
  arrange(desc(enrollment)) %>%
  select(nct, sponsor, title, phase_clean, status, condition, enrollment) %>%
  head(10)

top_enrollment_trials

The result was:

Rank	NCT ID	Sponsor	Phase	Status	Condition	Enrollment
1	NCT00744263	Pfizer	Phase 4	Completed	Pneumonia, Pneumococcal	84,496
2	NCT00090233	Merck	Phase 3	Completed	Rotavirus Infections	69,274
3	NCT00140673	GSK	Phase 3	Completed	Rotavirus Infections	63,227
4	NCT02251704	GSK	Not Available	Recruiting	Malaria	56,000
5	NCT02374450	GSK	Not Available	Recruiting	Malaria	52,192
6	NCT00344513	GSK	Phase 4	Completed	Heart Failure	50,000
7	NCT00709891	Roche	Not Available	Completed	Papilloma	47,208
8	NCT00753272	GSK	Phase 3	Completed	Influenza, Human	43,695
9	NCT00861380	GSK	Phase 3	Completed	Streptococcal Infections	41,188
10	NCT01427309	Sanofi	Phase 4	Completed	Influenza, Human	31,989

What These Large Trials Reveal

The largest enrollment values are mostly connected to:

• Vaccines
• Infectious diseases
• Public health studies
• Infant and elderly populations
• Screening programs
• Post-marketing or late-stage research

This explains why the enrollment distribution is skewed. A few very large studies have a strong effect on the mean.

Analyzing Clinical Trial Enrollment by Phase

---

Next, we analyzed whether some clinical trial phases usually enroll more participants than others.

enrollment_by_phase <- trial_clean %>%
  group_by(phase_clean) %>%
  summarise(
    trials = n(),
    median_enrollment = median(enrollment, na.rm = TRUE),
    mean_enrollment = round(mean(enrollment, na.rm = TRUE), 2),
    max_enrollment = max(enrollment, na.rm = TRUE)
  ) %>%
  arrange(desc(median_enrollment))

enrollment_by_phase

The result was:

Clinical Trial Phase	Trials	Median Enrollment	Mean Enrollment	Maximum Enrollment
Phase 3	4,887	351	796	69,274
Phase 2/Phase 3	139	249	566	8,031
Phase 4	2,015	159	518	84,496
Not Available	263	100	909	56,000
Phase 2	3,596	90.5	180	4,002
Phase 1/Phase 2	322	60	99.9	1,365
Early Phase 1	10	32	58.2	328
Phase 1	2,516	32	51.1	1,260

What This Table Shows

The table shows that later-stage trials generally have larger enrollment sizes than early-stage trials.

Phase 3 has the highest median enrollment at 351 participants.

Phase 1 and Early Phase 1 have the lowest median enrollment at 32 participants.

This matches clinical research logic: early-stage studies are usually smaller, while late-stage studies are usually larger.

Visualizing Median Enrollment by Clinical Trial Phase

---

We created a bar chart for median enrollment by phase.

ggplot(enrollment_by_phase, aes(x = reorder(phase_clean, median_enrollment), y = median_enrollment)) +
  geom_col() +
  coord_flip() +
  labs(
    title = "Median Enrollment by Clinical Trial Phase",
    x = "Clinical Trial Phase",
    y = "Median Enrollment"
  ) +
  theme_minimal()

The chart clearly shows that Phase 3 has the highest typical enrollment, while Phase 1 and Early Phase 1 have the lowest.

Main Insights from Enrollment by Phase

Finding	Interpretation
Phase 3 has the highest median enrollment	Late-stage trials are typically larger
Phase 1 has the lowest median enrollment	Early safety trials are usually small
Phase 2 is smaller than Phase 3	Enrollment increases as studies advance
Phase 4 has the highest maximum enrollment	Post-market studies can be very large
Not Available has a high mean	Missing phase records include some large studies
Median is better than mean	Enrollment is skewed by large outliers

Analyzing the Top Clinical Trial Sponsors

---

After phase, status, and enrollment analysis, we examined the sponsor column.

sponsor_table <- trial_clean %>%
  count(sponsor) %>%
  arrange(desc(n)) %>%
  head(10)

sponsor_table

The result was:

Rank	Sponsor	Number of Trials
1	GSK	2,473
2	Novartis	2,320
3	Pfizer	1,970
4	Merck	1,770
5	Sanofi	1,524
6	JNJ	1,143
7	Roche	1,095
8	Bayer	619
9	AbbVie	417
10	Gilead	417

What the Sponsor Table Shows

The sponsor table shows that a small group of large pharmaceutical companies account for a major share of clinical trial activity in this dataset.

GSK has the highest number of trials, followed by Novartis, Pfizer, Merck, and Sanofi.

This means trial sponsorship is highly concentrated among major pharmaceutical companies.

Calculating Sponsor Percentage Share

---

Next, we calculated what percentage of all trials each top sponsor represented.

sponsor_table <- trial_clean %>%
  count(sponsor) %>%
  arrange(desc(n)) %>%
  mutate(percentage = round((n / sum(n)) * 100, 2)) %>%
  head(10)

sponsor_table

The result was:

Rank	Sponsor	Number of Trials	Percentage
1	GSK	2,473	17.99%
2	Novartis	2,320	16.88%
3	Pfizer	1,970	14.33%
4	Merck	1,770	12.87%
5	Sanofi	1,524	11.09%
6	JNJ	1,143	8.31%
7	Roche	1,095	7.96%
8	Bayer	619	4.50%
9	AbbVie	417	3.03%
10	Gilead	417	3.03%

Top Sponsors Are Highly Concentrated

The top three sponsors are:

Sponsor	Percentage
GSK	17.99%
Novartis	16.88%
Pfizer	14.33%

Together, these three companies account for 49.20% of all trials.

The top five sponsors account for 73.16%.

This means clinical trial activity in this dataset is highly sponsor-concentrated.

Visualizing Top Clinical Trial Sponsors by Percentage Share

---

We created a sponsor percentage bar chart.

ggplot(sponsor_table, aes(x = reorder(sponsor, percentage), y = percentage)) +
  geom_col() +
  coord_flip() +
  labs(
    title = "Top Clinical Trial Sponsors by Percentage Share",
    x = "Sponsor",
    y = "Percentage of Trials"
  ) +
  theme_minimal()

The chart clearly shows that GSK, Novartis, Pfizer, Merck, and Sanofi dominate the dataset.

Analyzing the Top Medical Conditions in the Dataset

---

Next, we analyzed the condition column.

condition_table <- trial_clean %>%
  count(condition) %>%
  arrange(desc(n)) %>%
  head(10)

condition_table

The result was:

Rank	Medical Condition	Number of Trials
1	Diabetes Mellitus, Type 2	536
2	Breast Neoplasms	388
3	Pulmonary Disease, Chronic Obstructive	339
4	Hypertension	338
5	Asthma	334
6	Arthritis, Rheumatoid	333
7	Influenza, Human	324
8	Schizophrenia	292
9	Diabetes Mellitus	270
10	Alzheimer Disease	218

What the Condition Table Shows

The most common condition is Diabetes Mellitus, Type 2, with 536 trials.

Other common conditions include breast neoplasms, COPD, hypertension, asthma, rheumatoid arthritis, influenza, schizophrenia, diabetes mellitus, and Alzheimer disease.

This shows that the dataset covers many different disease areas.

---

Calculating Percentage Share of Top Medical Conditions

---

We calculated the percentage share of each top condition.

condition_table <- trial_clean %>%
  count(condition) %>%
  arrange(desc(n)) %>%
  mutate(percentage = round((n / sum(n)) * 100, 2)) %>%
  head(10)

condition_table

The result was:

Rank	Medical Condition	Number of Trials	Percentage
1	Diabetes Mellitus, Type 2	536	3.90%
2	Breast Neoplasms	388	2.82%
3	Pulmonary Disease, Chronic Obstructive	339	2.47%
4	Hypertension	338	2.46%
5	Asthma	334	2.43%
6	Arthritis, Rheumatoid	333	2.42%
7	Influenza, Human	324	2.36%
8	Schizophrenia	292	2.12%
9	Diabetes Mellitus	270	1.96%
10	Alzheimer Disease	218	1.59%

Dataset Is Medically Diverse

The top condition, Type 2 Diabetes, accounts for only 3.90% of all records.

This means no single medical condition dominates the dataset.

The dataset is sponsor-concentrated but medically diverse.

Visualizing the Top Medical Conditions

---

We created a bar chart for the top medical conditions.

ggplot(condition_table, aes(x = reorder(condition, percentage), y = percentage)) +
  geom_col() +
  coord_flip() +
  labs(
    title = "Top Medical Conditions in Clinical Trial Dataset",
    x = "Medical Condition",
    y = "Percentage of Trials"
  ) +
  theme_minimal()

The chart confirms that Type 2 Diabetes is the most frequent condition, but the dataset is spread across many disease areas.

Analyzing Clinical Trials by Start Year

---

Next, we analyzed the start_year column.

year_table <- trial_clean %>%
  count(start_year) %>%
  arrange(start_year)

year_table

The yearly table showed that trial records were very low before 2000, increased sharply in the early 2000s, peaked around 2006–2008, and then declined later.

The peak year was 2007, with 1,115 trials.

Main Insights from Yearly Trial Analysis

Finding	Interpretation
Earliest year is 1984	Dataset includes historical records
Trial counts are low before 2000	Earlier years are less represented
Counts rise sharply after 2000	Dataset activity increases strongly
Peak occurs in 2007	Highest trial count is 1,115
2006–2008 are strongest years	Main concentration of trial starts
2019–2020 are very low	Likely incomplete dataset coverage
Later years need caution	Avoid overclaiming from incomplete data

The main conclusion is that this dataset is heavily concentrated in the 2000s, especially around 2006 to 2008.

Visualizing Clinical Trials by Start Year

---

We created a line chart.

ggplot(year_table, aes(x = start_year, y = n)) +
  geom_line() +
  geom_point() +
  labs(
    title = "Clinical Trials by Start Year",
    x = "Start Year",
    y = "Number of Trials"
  ) +
  theme_minimal()

The chart shows:

• Very few records before 2000
• Rapid growth after 2000
• Peak around 2006–2008
• Decline after 2008
• Very low counts in 2019 and 2020

The drop in 2019 and 2020 should be interpreted carefully because it may reflect incomplete dataset coverage.

Analyzing Clinical Trials by Start Month

---

Next, we analyzed trial starts by month.

month_table <- trial_clean %>%
  count(start_month) %>%
  arrange(start_month)

month_table

The result was:

Month Number	Number of Trials
1	1,118
2	1,018
3	1,153
4	1,053
5	1,090
6	1,185
7	1,080
8	1,013
9	1,217
10	1,302
11	1,260
12	1,259

What the Monthly Table Shows

Clinical trial starts are fairly spread across all 12 months.

October has the highest number of trial starts, with 1,302 trials.

August has the lowest number, with 1,013 trials.

The difference between the highest and lowest month is 289 trials, which shows moderate variation but not extreme seasonality.

Converting Start Month Numbers into Month Names

Month numbers are not as reader-friendly as month names, so we converted them.

month_table <- month_table %>%
  mutate(
    month_name = month.abb[start_month],
    month_name = factor(month_name, levels = month.abb)
  )

month_table

The updated table was:

Month Number	Number of Trials	Month Name
1	1,118	Jan
2	1,018	Feb
3	1,153	Mar
4	1,053	Apr
5	1,090	May
6	1,185	Jun
7	1,080	Jul
8	1,013	Aug
9	1,217	Sep
10	1,302	Oct
11	1,260	Nov
12	1,259	Dec

Using factor() keeps the months in correct calendar order.

Visualizing Clinical Trials by Start Month

---

We created a monthly bar chart.

ggplot(month_table, aes(x = month_name, y = n)) +
  geom_col() +
  labs(
    title = "Clinical Trials by Start Month",
    x = "Start Month",
    y = "Number of Trials"
  ) +
  theme_minimal()

The chart confirms that trial starts are distributed across all months, with a slight increase in the final quarter.

Calculating Monthly Percentage Share

---

We calculated monthly percentages.

month_table <- month_table %>%
  mutate(percentage = round((n / sum(n)) * 100, 2))

month_table

The result was:

Month	Number of Trials	Percentage
Jan	1,118	8.13%
Feb	1,018	7.40%
Mar	1,153	8.39%
Apr	1,053	7.66%
May	1,090	7.93%
Jun	1,185	8.62%
Jul	1,080	7.86%
Aug	1,013	7.37%
Sep	1,217	8.85%
Oct	1,302	9.47%
Nov	1,260	9.16%
Dec	1,259	9.16%

Monthly Pattern

Each month contributes between 7.37% and 9.47% of all trials.

October has the highest share at 9.47%.

August has the lowest share at 7.37%.

The final quarter appears slightly higher, but the dataset does not show an extreme seasonal pattern.

Analyzing Clinical Trial Status by Phase

---

After analyzing phase and status separately, we compared them together.

status_by_phase <- trial_clean %>%
  count(phase_clean, status) %>%
  arrange(phase_clean, desc(n))

status_by_phase

This counts how many trials fall into each combination of clinical trial phase and trial status.

Why Status by Phase Analysis Matters

Phase tells us where a trial sits in the clinical research pipeline.

Status tells us what happened to the trial.

Together, they help us understand:

• Which phases have the most completed trials
• Which phases have more terminated studies
• Which phases have more recruiting or active studies
• Whether missing phase records still contain useful status information

Calculating Trial Status Percentage Within Each Phase

---

Raw counts can be misleading when phase groups have different sizes, so we calculated percentages within each phase.

status_by_phase_percent <- trial_clean %>%
  count(phase_clean, status) %>%
  group_by(phase_clean) %>%
  mutate(percentage = round((n / sum(n)) * 100, 2)) %>%
  arrange(phase_clean, desc(percentage))

print(status_by_phase_percent, n = Inf)

Completed Trials by Phase

Clinical Trial Phase	Completed Trials	Completion Percentage
Early Phase 1	10	100.00%
Not Available	218	82.90%
Phase 4	1,659	82.30%
Phase 1	1,977	78.60%
Phase 3	3,795	77.70%
Phase 2	2,646	73.60%
Phase 2/Phase 3	89	64.00%
Phase 1/Phase 2	174	54.00%

The main finding is that Completed is the largest status category in every phase.

Terminated Trials by Phase

Clinical Trial Phase	Terminated Trials	Termination Percentage
Phase 2/Phase 3	23	16.60%
Phase 2	449	12.50%
Phase 1/Phase 2	38	11.80%
Phase 4	183	9.08%
Phase 1	203	8.07%
Not Available	20	7.60%
Phase 3	369	7.55%

The highest termination percentage appears in Phase 2/Phase 3, followed by Phase 2.

This suggests that mid-stage and transition-stage trials may have relatively higher termination shares in this dataset.

Visualizing Clinical Trial Status Percentage by Phase

---

We created a stacked bar chart using the five main statuses:

status_by_phase_main <- status_by_phase_percent %>%
  filter(status %in% c(
    "Completed",
    "Terminated",
    "Recruiting",
    "Active, not recruiting",
    "Withdrawn"
  ))

ggplot(status_by_phase_main, aes(x = phase_clean, y = percentage, fill = status)) +
  geom_col() +
  coord_flip() +
  labs(
    title = "Clinical Trial Status Percentage by Phase",
    x = "Clinical Trial Phase",
    y = "Percentage Within Phase",
    fill = "Trial Status"
  ) +
  theme_minimal()

Main Insights from Status-by-Phase Analysis

Finding	Interpretation
Completed is the top status in every phase	Most trials reached completion
Phase 4 completion is 82.30%	Post-market trials have a high completion share
Phase 3 completion is 77.70%	Late-stage trials are strongly represented and often completed
Phase 2 termination is 12.50%	Mid-stage trials show meaningful early stopping
Phase 2/Phase 3 termination is 16.60%	Transition-stage trials have the highest termination share
Phase 1/Phase 2 recruiting is 18.00%	Hybrid early-stage studies include many ongoing trials
Withdrawn percentages are low	Few trials stop before enrollment

This chart gives one of the most advanced insights in the article because it connects two important clinical trial variables: phase and status.

Analyzing Sponsor Activity by Clinical Trial Phase

---

Next, we examined how the top sponsors are distributed across clinical trial phases.

sponsor_phase_table <- trial_clean %>%
  filter(sponsor %in% sponsor_table$sponsor) %>%
  count(sponsor, phase_clean) %>%
  arrange(sponsor, desc(n))

sponsor_phase_table

Largest Phase Category for Each Sponsor

Sponsor	Largest Phase Category	Number of Trials
GSK	Phase 1	723
Novartis	Phase 3	829
Pfizer	Phase 3	654
Merck	Phase 3	760
Sanofi	Phase 3	565
JNJ	Phase 3	451
Roche	Phase 3	394
Bayer	Phase 3	219
AbbVie	Phase 3	174
Gilead	Phase 2	152

This table shows that Phase 3 is the leading phase for most top sponsors.

---

Key Sponsor-by-Phase Insights

Finding	Interpretation
Phase 3 dominates most sponsors	Late-stage trials are central to the dataset
Novartis has the most Phase 3 trials	Strong late-stage sponsor presence
GSK has the most Phase 1 trials	Broad early-stage representation
Pfizer and Merck have high Phase 3 counts	Major late-stage research activity
Gilead leads slightly in Phase 2	More mid-stage focus in this dataset
Phase 4 appears across major sponsors	Post-market research is well represented

Visualizing Top Sponsors by Clinical Trial Phase

---

We created a heatmap.

ggplot(sponsor_phase_table, aes(x = phase_clean, y = sponsor, fill = n)) +
  geom_tile() +
  labs(
    title = "Top Sponsors by Clinical Trial Phase",
    x = "Clinical Trial Phase",
    y = "Sponsor",
    fill = "Number of Trials"
  ) +
  theme_minimal() +
  theme(
    axis.text.x = element_text(angle = 45, hjust = 1)
  )

A heatmap is useful because it compares two categorical variables at the same time:

Variable	Meaning
Sponsor	Company responsible for the clinical trial
Clinical trial phase	Stage of the trial
Fill intensity	Number of trials

The heatmap confirms that most top sponsors are heavily represented in Phase 2 and Phase 3 trials.

Analyzing Completed Clinical Trials by Sponsor

---

Next, we analyzed which sponsors had the highest number and percentage of completed trials.

completed_by_sponsor <- trial_clean %>%
  filter(sponsor %in% sponsor_table$sponsor) %>%
  count(sponsor, status) %>%
  group_by(sponsor) %>%
  mutate(percentage = round((n / sum(n)) * 100, 2)) %>%
  filter(status == "Completed") %>%
  arrange(desc(n))

completed_by_sponsor

The result was:

Rank	Sponsor	Status	Completed Trials	Completion Percentage
1	GSK	Completed	2,116	85.60%
2	Novartis	Completed	1,742	75.10%
3	Pfizer	Completed	1,473	74.80%
4	Merck	Completed	1,360	76.80%
5	Sanofi	Completed	1,232	80.80%
6	JNJ	Completed	868	75.90%
7	Roche	Completed	772	70.50%
8	Bayer	Completed	470	75.90%
9	Gilead	Completed	288	69.10%
10	AbbVie	Completed	247	59.20%

---

What the Completion Table Shows

GSK has the highest number of completed trials, with 2,116 completed studies.

GSK also has the highest completion percentage among the top sponsors, at 85.60%.

Sanofi also has a strong completion percentage of 80.80%.

AbbVie has the lowest completion percentage among the top 10 sponsors, at 59.20%.

This should be interpreted carefully because lower completion percentage may reflect more active, recruiting, or recently started trials rather than poor performance.

Step Visualizing Clinical Trial Completion Percentage by Sponsor

---

We created a completion percentage chart.

ggplot(completed_by_sponsor, aes(x = reorder(sponsor, percentage), y = percentage)) +
  geom_col() +
  coord_flip() +
  labs(
    title = "Clinical Trial Completion Percentage by Sponsor",
    x = "Sponsor",
    y = "Completion Percentage"
  ) +
  theme_minimal()

Main Insights from Sponsor Completion Chart

Finding	Interpretation
GSK has the highest completion percentage	Strongest completion share among top sponsors
Sanofi is second highest	Strong completion profile
Most sponsors fall between 69% and 77%	Similar completion patterns among major sponsors
AbbVie has the lowest completion percentage	More non-completed or active trials in this dataset
Completion percentage adds context beyond trial count	Count alone does not show sponsor outcome profile

Final Key Findings from the Clinical Trial Data Analysis

---

After completing the major analysis steps, we created a final summary table.

key_findings <- tibble(
  metric = c(
    "Total clinical trial records",
    "Most common phase",
    "Most common status",
    "Median enrollment",
    "Highest enrollment",
    "Top sponsor",
    "Top medical condition",
    "Peak start year",
    "Most common start month"
  ),
  finding = c(
    nrow(trial_clean),
    phase_table$phase_clean[1],
    status_table$status[1],
    median(trial_clean$enrollment),
    max(trial_clean$enrollment),
    sponsor_table$sponsor[1],
    condition_table$condition[1],
    year_table$start_year[which.max(year_table$n)],
    as.character(month_table$month_name[which.max(month_table$n)])
  )
)

key_findings

The result was:

Metric	Finding
Total clinical trial records	13,748
Most common phase	Phase 3
Most common status	Completed
Median enrollment	124
Highest enrollment	84,496
Top sponsor	GSK
Top medical condition	Diabetes Mellitus, Type 2
Peak start year	2007
Most common start month	Oct

Final Analysis Summary

---

This case study showed how R can be used to perform a complete exploratory analysis of a real-world clinical trial dataset.

Using R, we were able to:

• Clean the dataset
• Handle missing phase values
• Analyze clinical trial phases
• Analyze trial status
• Study enrollment distribution
• Compare sponsors
• Identify top medical conditions
• Explore yearly and monthly trends
• Compare trial status across phases
• Summarize the most important findings

The strongest findings were:

Area	Main Finding
Phase	Phase 3 trials were most common
Status	Completed trials dominated
Enrollment	Median enrollment was 124
Sponsor	GSK was the top sponsor
Condition	Type 2 diabetes was the top condition
Time	Peak start year was 2007
Month	October had the most starts

Conclusion: What We Learned from Clinical Trial Data Analysis Using R

---

This clinical trial data analysis using R case study showed how a real-world healthcare dataset can be cleaned, explored, visualized, and interpreted using modern R tools.

The purpose of this analysis was not to compare medicines, judge treatment effectiveness, or decide which clinical trial produced better patient outcomes. Instead, this project focused on understanding the structure and patterns inside the clinical trial dataset.

In simple terms, this analysis answered questions such as:

• What types of clinical trials are present in the dataset?
• Which clinical trial phases appear most often?
• What is the most common trial status?
• How large are the trials based on enrollment?
• Which sponsors appear most frequently?
• Which medical conditions are studied most often?
• When were most trials started?
• How are trial phases, sponsors, and statuses related?

This makes the project an example of exploratory clinical trial data analysis, not clinical medical evaluation.

Conclusion from the R Data Analysis Workflow

From a data analysis perspective, this case study demonstrated a complete beginner-to-intermediate workflow in R.

We started by importing the dataset into RStudio and checking its structure using commands such as:

summary(trial_data)
str(trial_data)
dim(trial_data)

After inspecting the raw data, we cleaned the dataset by standardizing column names, converting blank values into proper missing values, removing duplicate records, and creating new analysis-ready variables such as phase_clean.

This step was important because real-world datasets are rarely clean when they are first imported. The raw dataset contained missing values in important fields such as phase and title.

Instead of deleting useful records, we handled missing phase values transparently by labeling them as Not Available.

Using packages such as dplyr, ggplot2, janitor, and stringr, we created tables, percentages, summaries, bar charts, histograms, line charts, and heatmaps.

These tools allowed us to move from raw CSV data to meaningful insights.

How R Helped in This Analysis

Task	How R Helped
Data importing	Loaded the CSV dataset into RStudio
Data cleaning	Standardized columns and handled missing values
Descriptive analysis	Created summaries, counts, and percentages
Visualization	Produced charts for phase, status, enrollment, sponsors, and trends
Grouped analysis	Compared enrollment, status, and sponsors by phase
Reporting	Created structured tables suitable for a blog post or report

The analysis also showed why it is important to interpret data carefully.

For example, the enrollment column had a median of 124 but a mean of 440.78, showing that a few very large trials pulled the average upward.

Without comparing mean and median, the typical trial size could easily be misunderstood.

So, from a data analysis point of view, this project demonstrated how R can turn a large clinical trial dataset into a clear analytical story.

Conclusion from the Dataset Findings

From the dataset itself, several important patterns were found.

The dataset contained 13,748 clinical trial records, making it large enough for meaningful exploratory analysis.

The most common clinical trial phase was Phase 3, which represented the largest share of trial records. This showed that the dataset was strongly concentrated in late-stage clinical trial activity.

The most common trial status was Completed, with 76.87% of trials marked as completed. This means most records in the dataset represented trials that had reached completion, rather than trials that were still recruiting, suspended, withdrawn, or unknown.

The enrollment analysis showed that most clinical trials were small to medium-sized, while a few very large studies created extreme outliers.

The median enrollment was 124 participants, while the highest enrollment was 84,496 participants. This confirmed that trial enrollment was highly right-skewed.

The sponsor analysis showed that trial activity was highly concentrated among major pharmaceutical companies. GSK was the top sponsor, followed by Novartis, Pfizer, Merck, Sanofi, JNJ, Roche, Bayer, AbbVie, and Gilead.

The condition analysis showed a different pattern. Unlike sponsor activity, medical conditions were more diverse. The most common condition was Diabetes Mellitus, Type 2, but it represented only 3.90% of all records.

The time-based analysis showed that most trial records were concentrated in the 2000s, with the highest number of trial starts in 2007. The monthly analysis showed that trial starts were fairly balanced across the year, with October having the highest number of starts.

Important Limitation of This Analysis

This analysis should not be interpreted as medical evidence about which treatment works better, which drug is safer, or which trial improved patient outcomes.

The dataset used in this case study contains information about trial records, such as sponsor, phase, status, enrollment, start year, and condition.

It does not provide enough information to compare patient recovery, treatment effectiveness, survival outcomes, adverse events, or clinical success.

Therefore, this post should be understood as a study of clinical trial activity and structure, not a study of medical results.

We Analyzed	We Did Not Analyze
What trials were recorded	Which treatment worked best
Who sponsored the trials	Which medicine cured patients
When trials started	Patient recovery outcomes
Trial phase distribution	Drug safety or efficacy
Trial status distribution	Clinical superiority
Enrollment patterns	Individual patient health results
Medical condition frequency	Treatment success rates

This distinction is important because clinical trial data analysis can have different goals.

In this post, the goal was to understand the dataset and describe patterns in clinical trial activity.

Frequently Asked Questions About Clinical Trial Data Analysis Using R

---

What is clinical trial data analysis using R?

Clinical trial data analysis using R is the process of using the R programming language to clean, explore, summarize, and visualize clinical trial datasets.

In this case study, we used R to analyze trial phase, status, enrollment, sponsors, medical conditions, start year, and start month.

This analysis focused on understanding the structure of clinical trial records, not on judging whether one treatment or medicine is better than another.

What dataset was used in this clinical trial case study?

This case study used a clinical trial dataset containing 13,748 records.

The dataset included information such as trial ID, sponsor, title, summary, start year, start month, phase, enrollment, status, and medical condition.

The analysis was performed using R and RStudio.

What was the main goal of this analysis?

The main goal was to understand what clinical trials were recorded, when they started, who sponsored them, which phases they belonged to, how many participants they enrolled, and which conditions appeared most often.

This was an exploratory data analysis project, not a medical effectiveness study.

Did this analysis compare medicines or patient recovery outcomes?

No. This analysis did not compare medicines, treatments, patient recovery, survival outcomes, side effects, or clinical effectiveness.

The dataset was used to study clinical trial activity and structure, such as sponsor distribution, trial phases, trial status, enrollment size, and medical condition frequency.

Which R packages were used in this analysis?

The main R packages used in this project were:

Package	Purpose
tidyverse	General data analysis workflow
dplyr	Data cleaning and grouped summaries
ggplot2	Data visualization
readr	CSV importing
janitor	Cleaning column names
stringr	Text cleaning
lubridate	Date and time handling

These packages are commonly used in data analysis workflows with R.

What was the most common clinical trial phase in the dataset?

The most common clinical trial phase was Phase 3.

Phase 3 had 4,887 trials, representing 35.55% of the dataset.

This means late-stage confirmatory trials were the largest phase category in this dataset.

What was the most common clinical trial status?

The most common trial status was Completed.

Completed trials made up 76.87% of the dataset, with 10,568 trials marked as completed.

This shows that most records in the dataset represented trials that had reached completion.

Which sponsor had the most clinical trials?

The top sponsor in the dataset was GSK, with 2,473 trials, representing 17.99% of all trial records.

GSK also had the highest number of completed trials among the top sponsors.

Which medical condition appeared most often?

The most common medical condition was Diabetes Mellitus, Type 2.

It appeared in 536 trials, representing 3.90% of the dataset.

This shows that type 2 diabetes was the most frequent condition, but it did not dominate the entire dataset.

What was the median enrollment size?

The median enrollment was 124 participants.

This means half of the trials had enrollment below or equal to 124 participants, and half had enrollment above or equal to 124 participants.

The median was more useful than the mean because enrollment was highly skewed by a few very large studies.

What was the highest enrollment in the dataset?

The highest enrollment was 84,496 participants.

This very large value explains why the enrollment distribution was strongly right-skewed.

Most trials had much smaller enrollment numbers, but a few very large studies increased the mean enrollment.

Why was a log-scale histogram used for enrollment?

A log-scale histogram was used because enrollment values ranged from 0 to 84,496.

In the normal histogram, most trials were compressed near the left side because a few very large trials stretched the x-axis.

The log-scale chart made the distribution easier to interpret by spreading out smaller and medium-sized trial counts.

Which year had the most clinical trial starts?

The peak start year was 2007, with 1,115 trials.

The dataset showed strong trial activity around 2006 to 2008, while earlier years and the final years were less represented.

Which month had the most clinical trial starts?

The most common start month was October, with 1,302 trials, representing 9.47% of all records.

However, trial starts were fairly balanced across the year, so October was the highest month but not overwhelmingly dominant.

Why did we create a phase_clean variable?

We created phase_clean because the original phase column had missing values.

Instead of deleting those records, missing phase values were labeled as Not Available.

This allowed us to keep all trial records while making phase analysis clearer and more transparent.

Why should missing values not always be deleted?

Missing values should not always be deleted because some records may still contain useful information in other columns.

For example, a record with missing phase information may still have valid sponsor, status, enrollment, condition, and start year data.

In this analysis, we kept missing phase records and labeled them clearly instead of removing them.

What is the difference between raw data and cleaned data?

Raw data is the dataset exactly as imported from the CSV file. Cleaned data is the prepared version used for analysis.

Raw Data	Cleaned Data
Original imported file	Processed analysis-ready data
May contain blanks	Blanks converted to NA
Column names may be inconsistent	Column names standardized
May include duplicates	Duplicates checked/removed
Used for inspection	Used for final analysis

In this project, the cleaned dataset was stored as trial_clean.

What is the most important finding from this case study?

The most important finding is that the dataset is phase-concentrated, sponsor-concentrated, but medically diverse.

Phase 3 was the most common trial phase, GSK was the top sponsor, and completed trials dominated the status distribution.

However, medical conditions were spread across many disease areas rather than being dominated by one condition.

Is R good for clinical trial data analysis?

Yes. R is useful for clinical trial data analysis because it provides powerful tools for cleaning data, summarizing variables, calculating percentages, creating visualizations, and building reproducible workflows.

In this project, R helped transform a raw CSV dataset into tables, charts, and interpretable healthcare analytics insights.

Can this dataset be used for machine learning?

Yes, the dataset could be used for machine learning, but additional preprocessing would be needed.

Possible machine learning tasks include:

• Predicting trial completion status
• Classifying trials by phase
• Identifying sponsor patterns
• Analyzing trial summaries using text mining
• Predicting high-enrollment trials

However, this article focused on exploratory data analysis rather than predictive modeling.

What can be analyzed next?

Future analysis could include:

• Terminated trials by sponsor
• Trial status prediction
• Condition-specific trend analysis
• Sponsor comparison by disease area
• Text mining of trial summaries
• Machine learning classification
• Phase-wise enrollment modeling

These extensions would make the project more advanced and suitable for healthcare analytics or pharmaceutical data science portfolios.