Bio Statistics
Cancer A case–control study was performed early in the Nurses’ Health Study (NHS) to assess the possible association between oral contraceptive (OC) use and ovarian cancer. Forty seven ovarian cancer cases were identified at or before baseline (1976). For each case, 10 controls matched by year of birth and with intact ovaries at the time of the index woman’s diagnosis were randomly chosen from questionnaire respondents free from ovarian cancer. The data in Table 1 were presented.
Table 1 Duration of OC use by age at diagnosis among women with ovarian cancer and controls
|
|
|
|
Duration OC use |
|
|
Age at diagnosis |
|
Never |
<3 years |
3+ years |
|
Under 35 |
Case |
9 |
2 |
0 |
|
|
Control |
55 |
42 |
12 |
|
35–44 |
Case |
13 |
2 |
4 |
|
|
Control |
127 |
27 |
30 |
|
45+ |
Case |
12 |
3 |
2 |
|
|
Control |
129 |
18 |
23 |
1 Use logistic regression methods to assess whether there is an association between ovarian cancer risk and duration of OC use while controlling for age. Provide a two-sided p-value. Assume that the average duration of use in the < 3 years group = 1.5 years and in the 3+ years group = 4 years. Also, provide an estimate of the OR relating ovarian cancer risk per year of use of OCs and a 95% CI.
2 Use logistic regression methods to assess whether there is an association between ever use of OCs and ovarian cancer risk, while controlling for age. Also, provide an estimate of the OR and a 95% CI about this estimate.
Cancer
The data file BLOOD.DAT (www.cengagebrain.com) contains data from a case–control study assessing several plasma risk factors for breast cancer. The women were matched approximately by age at the blood draw, fasting status and, if possible, current PMH use at the time of the blood draw. There was 1 case and either 1 or 2 controls per matched set, although some of the matched sets are incomplete due to missing data. The matching variable is matchid. Use logistic regression methods to assess the association between testosterone and breast cancer risk after controlling for age at the blood draw and current PMH use and taking the matching into account.
Perform the analysis in two ways:
3 Treat testosterone as a continuous variable (suitably transformed if necessary).
4 Treat testosterone as a categorical variable in quartiles, with the 1st quartile as the reference group.(hint: You may want to do a recoding or transformation in similar way as we have done before)
5 Discuss your results from Problems 3 and 4.