Breast on individual risk factor profiles, using genetic risk

Breast cancer is the most
common cancer in women, and the second leading cause of cancer deaths.

Currently, the average risk of a woman in both Switzerland and the United
States developing breast cancer sometime in her life is about 12%, which means
about 1 in 8 women will develop breast cancer(1-3). Since 2009 the United States Preventive Services Task Force (USPSTF)
has recommend biennial screening for women age 50 to 74 years (4). In 2013 Switzerland has also adopted a national strategy against
cancer, recommending systematic screening for women over 50(1, 5). However, current screening program in both
countries has saved 1-2 women’s lives for every 1000 screened. Meanwhile beside
the cost effectiveness of programs, for the one breast-cancer death prevented
woman over a 10-year course of annual screening beginning at 50 years of age,
490 to 670 women are likely to have a false positive mammogram with repeat
examination; 70 to 100, an unnecessary biopsy; and 3 to 14, an over diagnosed breast
cancer that would never have become clinically apparent(6). Therefore more effort must be made to improve the
quality of mammography screening to reduce unnecessary investigations and offer
interventions to the “high risk” population.

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

 

Several approaches have been
invented, such as tailoring screening frequency based on individual risk factor
profiles, using genetic risk thresholds rather than age and so on(7-10). Many breast risk assessment tools (e.g. Gail model, BOADICEA modelCM1 , Tyrer-Cuzick model and so on) were developed since 1980s
aim to classify the high risk group in population based on risk factors of
breast cancer or characteristics of high risk population based on large scale cohorts
data. They were combined into clinical guidelines to identify women who were at
elevated risk, then recommended them to have high-risk treatment or
risk reducing medication(11). Although some of these models have been well
calibrated and validated in a multiracial and multiethnic population with large
sample sizes, their c-statistic stayed in the range from 0.53 to 0.64(12-15). They can predict almost precisely how many women will develop breast
cancer in a given population, but cannot predict accurately who will develop
the disease. There are about 36% to 47% chance individuals at high risk
of breast cancer who
fail to
be identified by these tools, while
some individuals not at
risk are
given preventive treatment unnecessarily. Because most current breast cancer risk assessment models make an implicit assumption that each risk factor is related in a linear
fashion to breast cancer appearance. Such models may
thus oversimplify complex relationships which include large numbers of risk factors with non-linear interactions(14). Approaches
with high accuracy, better incorporating multiple risk factors, determining
more nuanced relationships
between risk factors and outcomes are needed in breast cancer personalized risk
prediction.

 

Machine learning (ML) CM2 offers an alternative approach
to standard prediction modeling
that may
address current
limitations(16). ML was developed from the study of pattern recognition
and computational learning. This relies on a computer program to learn all complex and
unknown interactions between variables
by
minimizing the error between predicted and observed outcomes(17). ML has been shown
better prediction accuracy and reliability in different cancer prognostic
prediction and survival prediction(18-21). Until this day, there has been no investigation applying machine learning internationally
for individual breast cancer risk prediction and
compared its predictive
accuracy with existing models. The aim of this study was to evaluate whether machine learning can improve accuracy of personalized breast cancer risk prediction
by comparing to Gail model and BOADICEA model. We also researched to determine which machine-learning algorithm has the highest predictive accuracy.

 CM1Do
you feel necessary that I introduce these two models I will be comparing with
her? I put them into method section, explained how I got the predictions from
Gail and BOADICEA

 CM2 I
think I need to introduce ML better here, any idea?

x

Hi!
I'm Josephine!

Would you like to get a custom essay? How about receiving a customized one?

Check it out