Applied Statistical Methods

HSRP 734

Summer 2008

Homework 5

Due: 7/24/2008

General instructions: 1) you may discuss any and all portions of the assignment with other members of the class. However, the homework you turn in must be your own. 2) For problems that require a statistical package, you can use any statistical software, but must do your own programming. 3) A final answer is not sufficient. Show all your work. Please provide relevant SAS output with clearly indicated answers to the questions.

Q1. Dr. Lobo conducted a follow-up study of 301 female patients who were diagnosed with breast cancer during the period January 1, 2001 – December 31, 2005. The day after each subject’s diagnosis, she ascertained a complete family history of breast cancer. Subjects were followed for death through May 31, 2007.

Let XA = age of patient in years at time of diagnosis.

XM = 1 if the subject’s mother had developed breast cancer before the subject’s diagnosis

0 if the subject’s mother had NOT developed breast cancer before the subject’s diagnosis

She determined that the following Cox regression model fit her censored survival data well:

Q1a. According to this model, what is the age-at-diagnosis adjusted relative risk of death comparing subjects with a maternal history of breast cancer at diagnosis to subjects with no maternal history of breast cancer at diagnosis, both at the same follow-up time?

Q1b. According to this model, what is the relative risk of death comparing a woman who was 63 at diagnosis and who had a maternal history of breast cancer at diagnosis to a woman who was 45 at diagnosis who did not have a maternal history of breast cancer at diagnosis, both at the same follow-up time?

Q1c. Why is the bolded phrase in the previous question important?

Q1d. According to this model, what is the relative risk of death comparing a woman who was 63 at diagnosis and who had a maternal history of breast cancer at diagnosis 20 months after diagnosis to a woman who was 45 at diagnosis who also had a maternal history of breast cancer at diagnosis 30 months after diagnosis?

Q2. Use the SAS dataset hw5Data from the class website to answer the following questions. These are leukemia patients who underwent bone marrow transfers either allogeneic or autologous. Measures of their pre-transplant body iron burden were measured

Variable Label

AGE_AT_BMT AGE AT BMT

Ferritin Ferritin level

ID ID

Iron Serum iron level

Female M/F 1=female 0= male

Percent_saturation Percent saturation

Survival_in_days Survival in days

Transferrin Transferrin level

bmt_allo bone marrow transfer type 1=allogenic 0=autologous

aml_diag diagnosis 1=AML 0=other

status status 0=censored

ferritin_diagnosis ferritin * aml_diag

Q2a Fit a Cox PH model with covariates M/F, age, bone marrow transfer type, diagnosis and ferritin level. Which factors contribute significantly to the model?

Q2b. Provide hazard ratio estimates and 95% confidence intervals and interpretations for all five variables in the model.

Q2c. What is the adjusted relative risk of death (and 95% CI) comparing a subject with a ferritin level of 500 to a subject with a ferritin level of 1000, both at the same follow-up time?

Q2d. A new model with one additional term will be necessary to answer this question. Adjust for the same factors as before, does diagnosis modify the risk of death for various levels of ferritin level?

Q2e – extra credit – but you must try. Using the model in d) what is the adjusted relative risk of death (and 95% CI) comparing a subject with a ferritin level of 500 to a subject with a ferritin level of 1000, both at the same follow-up time and both with a diagnosis of AML? What is the adjusted relative risk of death (and 95% CI) comparing a subject with a ferritin level of 500 to a subject with a ferritin level of 1000, both at the same follow-up time and both with a diagnosis of other?