Goal

This analysis will investigate potential factors towards a patient’s survival of breast cancer.

Such as:

Protein Evaluation Levels
Type of Surgery
Stage of Cancer
Patient’s Age

Understanding what may influence survival could help more people struggling with breast cancer.

Understanding the Data

Data was obtained from Kaggle

Libraries

# Loading libraries necessary for analysis 
library(tidyverse) 
library(janitor)
library(ggpubr) # Obtain function needed for Correlation Coefficient

View the Data

View the first few rows to see how the data was organized

cancer_data <- read_csv('Breast_Cancer_Data.csv')

TABLE 1.1: Glimpse of the Data
Age	Gender	Protein1	Protein2	Protein3	Protein4	Tumour_Stage	Histology	Surgery_type	Surgery_Date	Last_Visit_Date	Patient_Status	Surgery_Year
36	FEMALE	0.08	0.43	0.55	0.27	III	Infiltrating Ductal Carcinoma	Modified Radical Mastectomy	2017-01-15	2017-06-19	Alive	2017
56	FEMALE	0.35	-0.21	-0.19	0.12	II	Infiltrating Ductal Carcinoma	Modified Radical Mastectomy	2017-01-25	2017-07-12	Alive	2017
53	FEMALE	-0.07	1.42	-0.36	0.39	II	Infiltrating Ductal Carcinoma	Simple Mastectomy	2017-02-04	2018-02-07	Alive	2017
40	FEMALE	-0.57	1.27	-0.29	0.19	II	Infiltrating Lobular Carcinoma	Other	2017-02-14	2017-12-15	Alive	2017
75	FEMALE	0.17	0.03	-0.16	0.67	I	Infiltrating Lobular Carcinoma	Simple Mastectomy	2017-02-24	2017-04-05	Alive	2017
41	FEMALE	0.14	1.08	0.21	0.97	I	Infiltrating Lobular Carcinoma	Simple Mastectomy	2017-03-06	2019-08-09	Alive	2017

Data’s Structure

View the column names

colnames(cancer_data)

##  [1] "Age"             "Gender"          "Protein1"        "Protein2"       
##  [5] "Protein3"        "Protein4"        "Tumour_Stage"    "Histology"      
##  [9] "Surgery_type"    "Surgery_Date"    "Last_Visit_Date" "Patient_Status" 
## [13] "Surgery_Year"

There are 334 different patients with 13 columns of patient information

glimpse(cancer_data)

## Rows: 334
## Columns: 13
## $ Age             <dbl> 36, 56, 53, 40, 75, 41, 62, 45, 62, 50, 61, 43, 74, 56…
## $ Gender          <chr> "FEMALE", "FEMALE", "FEMALE", "FEMALE", "FEMALE", "FEM…
## $ Protein1        <dbl> 0.080353, 0.345090, -0.069535, -0.565700, 0.171640, 0.…
## $ Protein2        <dbl> 0.426380, -0.211470, 1.418300, 1.266800, 0.029656, 1.0…
## $ Protein3        <dbl> 0.547150, -0.193040, -0.361050, -0.293460, -0.158900, …
## $ Protein4        <dbl> 0.273680, 0.124270, 0.391580, 0.193950, 0.674710, 0.97…
## $ Tumour_Stage    <chr> "III", "II", "II", "II", "I", "I", "II", "I", "II", "I…
## $ Histology       <chr> "Infiltrating Ductal Carcinoma", "Infiltrating Ductal …
## $ Surgery_type    <chr> "Modified Radical Mastectomy", "Modified Radical Maste…
## $ Surgery_Date    <date> 2017-01-15, 2017-01-25, 2017-02-04, 2017-02-14, 2017-…
## $ Last_Visit_Date <date> 2017-06-19, 2017-07-12, 2018-02-07, 2017-12-15, 2017-…
## $ Patient_Status  <chr> "Alive", "Alive", "Alive", "Alive", "Alive", "Alive", …
## $ Surgery_Year    <dbl> 2017, 2017, 2017, 2017, 2017, 2017, 2017, 2017, 2017, …

Comparing Data Statistics

The following tables highlight:

The amount of patients
Average expression levels of Proteins 1 - 4
Percentage of Patients alive
Amount of patients whose survival status is unknown
Percentage highlighting the amount of cases of stages 1-3
The following table shows the statistics between all 334 patients:

TABLE 2.1: Data Set Statistics
Patients	Protein 1	Protein 2	Protein 3	Protein 4	Survival Rate	Unknown	Stage I	Stage II	Stage III
334	-0.03	0.95	-0.09	0.01	76.35%	13	19%	57%	24%

Splitting the data by the year a patient had their surgery:

TABLE 2.2: Yearly Statistics
Surgery Year	Patients	Protein 1	Protein 2	Protein 3	Protein 4	Survival Rate	Unknown	Stage I	Stage II	Stage III
2017	61	0.08	0.94	-0.09	0.16	75.41%	3	25%	57%	18%
2018	152	-0.07	0.91	-0.14	-0.10	77.63%	4	18%	56%	26%
2019	121	-0.03	1.00	-0.02	0.07	75.21%	6	17%	57%	26%

Result:

The survival rate is consistent throughout the data
- About 75%-77%
2018-2019 had similar percentages across stages 1-3
- 2017’s difference possibly stems from smaller pool of patients
The Patients whose status is unknown is spread unevenly
- Lack of information in data set

Future analysis needed to grasp the difference between the Protein expression levels throughout the years.

Comparing Protein Expression Levels

The following graphs’ axis are the patient’s expression levels for two different proteins

Expression Levels: State how proteins are maintained and changed
Goal: To identify if there is a relationship between 2 Proteins that can help identify a patient’s survival

Protein 1 vs 2

Chart

Protein 1 and Protein 2 have no significant positive correlation
- R Value: 0.24
  - The closer to 1 or -1 , the stronger the correlation
- The p value indicates this is a significantly insignificant result
Although there are no deceased patients when both the proteins are low, the lack of a significant correlation makes these findings unreliable.

Code

ggplot(data = cancer_data,mapping= aes(x = protein1,y = protein2)) +
  geom_point(aes(color = patient_status)) +
  geom_smooth(method = lm, se = FALSE) +
  stat_cor(method = "pearson", label.x = -1, label.y = 3) +
  labs(title = "Expression Levels Between Protein 1 & Protein 2",
       x = "Protein 1",
       y = "Protein 2", 
       color = "Patient Status") +
  theme(plot.background = element_rect(fill = '#FBD7CD'),
        legend.background = element_rect(fill = "#FBD7CD"))

Protein 1 vs 3

Chart

Protein 1 and Protein 3 have no significant negative correlation
- R Value: -0.13
- The p value indicates this is a significantly insignificant result
Although there are no deceased patients when protein 1 is low and protein 3 is high, the lack of a significant correlation makes these findings unreliable.

Code

ggplot(data = cancer_data,mapping = aes(x = protein1,y=protein3)) +
  geom_point(aes(color = patient_status)) +
  geom_smooth(method = lm, se = FALSE) +
  stat_cor(method = "pearson", label.x = -1, label.y = 3) +
  labs(title = "Expression Levels Between Protein 1 & Protein 3",
       x = "Protein 1",
       y = "Protein 3", 
       color = "Patient Status") +
  theme(plot.background = element_rect(fill = '#FBD7CD'),
        legend.background = element_rect(fill = "#FBD7CD"))

Protein 1 vs 4

Chart

Protein 1 and Protein 3 have an insignificant positive correlation
- R Value: 0.28
- The p value indicates this is a significantly insignificant result
Although there are no deceased patients when both proteins are low, the lack of a significant correlation makes these findings unreliable.

Code

ggplot(data = cancer_data, mapping = aes(x = protein1,y=protein4)) +
  geom_point(aes(color = patient_status)) +
  geom_smooth(method = lm, se = FALSE) +
  stat_cor(method = "pearson", label.x = -1, label.y = 3) +
  labs(title = "Expression Levels Between Protein 1 & Protein 4",
       x = "Protein 1",
       y = "Protein 4", 
       color = "Patient Status") +
  theme(plot.background = element_rect(fill = '#FBD7CD'),
        legend.background = element_rect(fill = "#FBD7CD"))

Protein 2 vs 3

Chart

Protein 2 and Protein 3 have an insignificant negative correlation
- R Value: -0.42
- The p value indicates this is a significantly insignificant result
The Patient Status is scattered throughout the plot

Code

ggplot(data = cancer_data, mapping = aes(x = protein2,y=protein3)) +  
  geom_point(aes(color = patient_status)) +
  geom_smooth(method = lm, se = FALSE) +
  stat_cor(method = "pearson", label.x = -1, label.y = 3) +
  labs(title = "Expression Levels Between Protein 2 & Protein 3",
       x = "Protein 2",
       y = "Protein 3", 
       color = "Patient Status") +
  theme(plot.background = element_rect(fill = '#FBD7CD'),
        legend.background = element_rect(fill = "#FBD7CD"))

Protein 2 vs 4

Chart

Protein 2 and Protein 4 have an insignificant positive correlation
- R Value: 0.088
The p value indicates this is unreliable result
- Some form of bias in data

Code

ggplot(data = cancer_data, mapping = aes(x = protein2,y=protein4)) +
  geom_point(aes(color = patient_status)) +
  geom_smooth(method = lm, se = FALSE) +
  stat_cor(method = "pearson", label.x = -1, label.y = 3) +
  labs(title = "Expression Levels Between Protein 2 & Protein 4",
       x = "Protein 2",
       y = "Protein 4", 
       color = "Patient Status") +
  theme(plot.background = element_rect(fill = '#FBD7CD'),
        legend.background = element_rect(fill = "#FBD7CD"))

Protein 3 vs 4

Chart

Protein 3 and Protein 4 have an insignificant positive correlation
- R Value: 0.065
The p value indicates this is unreliable result
- Some form of bias in data

Code

ggplot(data = cancer_data, mapping = aes(x = protein3,y=protein4)) +
  geom_point(aes(color = patient_status)) +
  geom_smooth(method = lm, se = FALSE) +
  stat_cor(method = "pearson", label.x = -1, label.y = 3) +
  labs(title = "Expression Levels Between Protein 3 & Protein 4",
       x = "Protein 3",
       y = "Protein 4", 
       color = "Patient Status") +
  theme(plot.background = element_rect(fill = '#FBD7CD'),
        legend.background = element_rect(fill = "#FBD7CD"))

Conclusion

The Protein Expression levels hold no strong correlation with each other
- Cannot be used to predict a patient’s expression levels
- Unreliable patterns such as:
  - No deceased patients when protein 1 and protein 2 are low
  - Could be sample bias
- No discernible pattern with patient survival

Surgery Type

Next is an attempt to identify whether the type of surgery patient their survival later in life

Try to see if the cancer’s stage affects the results

Chart

Code

# Grouped  bar chart showing the patient status for each surgery type divided by cancer stage
ggplot(data = cancer_data, aes(x = tumour_stage))+
  geom_bar(aes(fill = patient_status),position = 'dodge') + 
  geom_text(aes(label = after_stat(count),group = patient_status), stat ="count", 
            size = 3, position = position_dodge(0.8), vjust=-0.5) +
  facet_wrap(~surgery_type,scales = 'free')  + 
  ylim(0,45) +
  labs(title = "Surgeries Performed at Different Stages of Cancer",
       subtitle = "Split By Patient Survival",
       x = "Tumor Stage",
       y = "Amount of Patients", 
       fill = "Patient Status") +
  theme(plot.background = element_rect(fill = '#FBD7CD'),
        legend.background = element_rect(fill = "#FBD7CD"))

Stage 1:

Lumpectomy had the most patients alive and tied for the least deceased
Modified Radical Mastectomy most concerning when 5 alive and 2 deceased
- Seems unreliable

Stage 2:

Simple Mastectomy had 38 alive patients, while having the least deceased (5)
Modified Radical Mastectomy had same amount of alive patients, but almost twice as many later deceased (9)
Lumpectomy had least patients alive (29), while having concerning amount of deaths (7)
Other types of surgeries are the most concerning with 17 patients dead and 39 alive

Stage 3:

Lumpectomy remarkably had no deaths, with 8 alive
Simple Mastectomy is unreliable with 5 alive and 4 dead
Modified Radical Mastectomy has a good ratio for an advanced cancer stage
- 29 Alive, 9 deceased

Conclusion:

The type of surgery seems to correlate with the likelihood a patient will live years after surgery.

Some surgeries perform far better in different stages:

Lumpectomy is the most reliable for stage 1, and great for stage 3
Simple Mastectomy has favorable results for stage 2
Modified Radical Mastectomy has consistent patterns for stage 2 & 3

Limitation:

The data set doesn’t specify if the patient’s cause of death is related to the cancer or the result of something else.

Further analysis is needed for clarification

Cancer Stages Spread Across Ages

The goal is to identify how patient’s ages spread in the data set.

What ages have the most instances of cancer
How different stages of cancer is spread by age
Identify patterns between patients that are alive or deceased

Chart

Code

cancer_data %>% 
  filter(patient_status != 'Unknown') %>% 
  ggplot() +
    geom_histogram(aes(x = age, fill = tumour_stage),position = position_dodge(4),
                    binwidth = 5) +
    labs(title = "Age Range of Patients",
         subtitle = "Split By Tumor Stage",
         x = "Age",
         y = "Amount of Patients", 
         fill = "Tumor Stage",
         caption = "Bin Range of 5") +
    theme(plot.background = element_rect(fill = '#FBD7CD'),
          legend.background = element_rect(fill = "#FBD7CD"),
          plot.caption.position = 'plot') +
    facet_wrap(~patient_status)

There is a notable bell curve in the data:

Most of the data is between ages 40-80 for both groups

Patients that are alive:

Most cases between ages 40-80 see Stage 2 as the most common diagnosed
- Stage 1 and 3 are both prominent between 40-80 as well
Cases rise as patients grow closer to 60, but decrease the older patients are

Patients that are deceased:

Very rarely did a patient die from Stage 1, regardless of age.
Most cases of death were patients diagnosed with Stage 2 or 3
- Stage 3 being prominent between late 40s and late 50s
Rise in cases between late 40s and 70s, but a significant decrease afterwards.

Patient Survival By Surgery Year

Next, it is to investigate whether the year a patient had their surgery would affect their chances of a future.

Chart

Although 2018 seemed to perform the best, there are a few things to consider:

There are far more patients in the data set whose surgery took place in 2018
- Table 2.2:
  - 2017 : 61 Patients
  - 2018 : 152 Patients
  - 2019 : 121 Patients
The survival rate is almost the same, regardless of the amount of patients
- Table 2.2:
  - 2018 : 75.41%
  - 2018 : 77.63%
  - 2018 : 75.21%

It is unlikely the year a patient had their surgery would affect their chances of survival.

Code

cancer_data %>% 
  group_by(surgery_year,patient_status) %>% 
  summarize(count = n()) %>% 
  ggplot(mapping =aes(x = surgery_year, y = patient_status,fill = count)) +
    geom_tile(color = 'black', lwd = 1.5,) +
    geom_text(aes(label = count), color = 'black',size = 6) +
    coord_fixed() +
  scale_fill_distiller(palette = 'OrRd',direction = +1) +
  labs(title = "Current Patient Survival Split By Year",
       y = "Patient Status",
       x = "Year of Surgery", 
       fill = "Number of Patients",
       caption = 'Some patients whereabouts are unknown') +
  theme(plot.background = element_rect(fill = '#FBD7CD'),
        panel.background = element_rect(fill = '#FBD7CD'),
        legend.background = element_rect(fill = "#FBD7CD"),
        plot.caption.position = 'plot')

Breast Cancer Dashboard

Summarizes key points discussed:

Stages of cancer between ages
Scatter plots for protein expression levels
Survival rate of different surgeries

Additionally, filters allow for a more in-depth look at the data

Most filters are connected between all three charts
Age Range applies to Different Stages of Cancer Split By Age & Protein Expression Levels

Allows for user to specify:

Age range
Stage of Cancer
Type of Breast Cancer
Patient’s Status (Alive, Dead, or Unknown)

The Histology filter helps further identify any patterns among the different surgeries

Stage 1

Lumpectomy had two patients deceased for Stage 1
- Both are from an Infiltrating Ductal Carcinoma
- 3-0 for Infiltrating Lobular Carcinoma & 2-0 Mucinous Carcinoma
Modified Radical Masectomy performance dependent on type of cancer:
- Infiltrating Ductal Carcinoma: 4-2
- Mucinous Carcinoma : 1-0

Stage 2

Previously stated Simple Mastectomy had 38 alive patients, while having the least deceased (5)
- Filter showed it performs best against Infiltrating Ductal Carcinoma (21-4) but not as impressive against Infiltrating Lobular Carcinoma
  - Simple Mastectomy : 7-1
  - Modified Radical Mastectomy : 13-1
  - Lumpectomy : 7-1
Lumpectomy was shown to have the least patients alive (29), while having concerning amount of deaths (7)
- Yet filters highlight its good performance against Infiltrating Ductal Carcinoma (21-4) & Infiltrating Lobular Carcinoma (7-1)
- Performed poorly against Mucinous Carcinoma (1-2)

Stage 3

Lumpectomy had no deaths against Infiltrating Ductal Carcinoma & Infiltrating Lobular Carcinoma
- No patients had Mucinous Carcinoma for Stage 3
Simple Mastectomy is still unreliable against Infiltrating Ductal Carcinoma (4-4)
- Although against Infiltrating Lobular Carcinoma it had no deaths (1-0)
Modified Radical Mastectomy is great against Infiltrating Ductal Carcinoma (23-4)
- Conscerning against Infiltrating Lobular Carcinoma (6-5)

Conclusion

Utilizing filters allowed for a deeper analysis on the performances of different surgeries

Highlights which surgeries were better for which type of breast cancer

These factors seem to influence a patient’s chances of survival post-surgery:

Different stages of cancer
The type of Breast Cancer
The type of Surgery

Final Thoughts

It is unlikely the year a patient had their surgery would affect their chances of survival.
- The survival rate is consistent
  - About 75%-77%
The Protein Expression levels hold no strong correlation with each other
- Cannot be used to predict a patient’s expression levels
- No discernible pattern with patient survival
Those with cancer are likely to be between ages 40-80
- Stage 2 is the most common
- Cases rise as patients grow closer to 60, but decrease the older patients are
- Very rarely did a patient die from Stage 1, regardless of age.
Factors seemingly influence a patient’s chances of survival post-surgery:
- Different stages of cancer
- The type of Breast Cancer
- The type of Surgery

Limitations:

Unspecific cause of death
Lack of information on the “Unknown” Patients
Possible sample bias in data set
Unequal spread of type of cancer patients had

Further analysis needed to ensure these limitations didn’t impact the results.

Breast Cancer Analysis

Steven Hernandez

2023-04-16

Goal

Understanding the Data

Libraries

View the Data

Data’s Structure

Comparing Data Statistics

Comparing Protein Expression Levels

Protein 1 vs 2

Chart

Code

Protein 1 vs 3

Chart

Code

Protein 1 vs 4

Chart

Code

Protein 2 vs 3

Chart

Code

Protein 2 vs 4

Chart

Code

Protein 3 vs 4

Chart

Code

Conclusion

Surgery Type

Chart

Code

Cancer Stages Spread Across Ages

Chart

Code

Patient Survival By Surgery Year

Chart

Code

Breast Cancer Dashboard

Stage 1

Stage 2

Stage 3

Conclusion

Final Thoughts