Oct 2, 2014
BIOST 536
HW 1
1. I took a stratified analysis approach.
Select
Tx: idarubicin (N = 65)
Characteristics
Tx: daunorubicin (N =
65)
n
%
n
%
30
46.2
35
53.9
<30
21
32.3
19
29.2
30-39
17
26.2
13
20.0
40-49
12
18.5
12
18.5
>49
15
23.1
21
32.3
<60
2
3.1
3
4.6
60
4
6.2
5
7.7
70
11
16.9
6
9.2
80
25
38.5
27
41.5
>80
23
35.4
24
36.9
8
12.3
6
9.2
1
13
20.0
6
9.2
2
15
23.1
15
23.1
3
11
16.9
9
13.8
4
8
12.3
12
18.5
Male
Age
Karnofsky
score
Bone marrow
transplant
FAB
classification
2.
5
12
18.5
13
20.0
6
2
3.1
1
1.5
missing
4
6.2
9
13.8
I calculated a risk difference and relative risk from a standard 2x2 table
along with a chi-squared test to assess the likelihood that such a difference
would be expected by chance.
In the idarubicin treatment arm, 51 out of 65 (78.5%) participants
experienced complete remission compared with 38 out of 65 (58.5%) in the
daunorubicin arm. The risk difference of 20.0% is beyond what would be
expected by chance in the absence of a true treatment effect (95% CI: 4.40%
– 35.6%, p = 0.014). Patients treated with idarubicin were 1.34 times as likely
to experience complete remission (95% CI: 1.05 – 1.71).
3. Since the data was collected in a randomized trial, it is unlikely that there is
substantial confounding. As seen in the table from question one are slight
differences in the proportion of males and females in each treatment arm. To
assess if there may be residual confounding, I compared the sex-adjusted
Mantel-Haenszel RRs that I calculated for question 4. The sex-adjusted
Mantel-Haenszel RR for idarubicin treatment compared to daunorubicin
treatment is 1.31. While there is no strict test for confounding, when
comparing the sex-adjusted RR to the crude RR of 1.34 they are similar
enough to say that any residual confounding due to sex is minor.
4. As described in question 3, the sex-adjusted RR (calculated using MantelHaenszel stratification methods) for idarubicin treatment compared to
daunorubicin treatment is 1.31.
5. While males undergoing idarubicin treatment appear to experience
complete remission more frequently than those undergoing daunorubicin
treatment, the confidence interval is too wide to say with any certainty (RR =
1.44, 95% CI: 0.953 – 2.18).
6. While females undergoing idarubicin treatment appear to experience
complete remission somewhat more frequently than those undergoing
daunorubicin treatment, the study was underpowered to answer this question
and the confidence interval is far too wide to say with any certainty (RR =
1.22, 95% CI: 0.934 – 1.60).
7. I created multiple 2x2 tables stratified by sex and calculated RRs for each.
Sex may modify the effect of the treatment of idarubicin compared with
daunorubicin though the study is underpowered to suitably examine this
question (p = .50). Males undergoing idarubicin treatment are 1.44 times as
likely to experience complete remission (95% CI: 0.953 – 2.18) whereas
females are 1.22 times as likely (95% CI: 0.934 – 1.60).
8. Yes, idarubicin treatment is associated with more frequent induction of
complete remission (RR = 1.34, 95% CI: 1.05 – 1.71). Sex may modify the
effect with males undergoing idarubicin treatment being 1.44 times as likely
to experience complete remission (95% CI: 0.953 – 2.18) and females being
1.22 times as likely (95% CI: 0.934 – 1.60).
9.
Observed probability of complete remission
Tx: Idarubicin (N = 65)
Sex
n
Male
21/30
Female 30/35
Tx: Daunorubicin (N = 65)
%
n
%
70.0% 17/35
48.6%
85.7% 21/30
70.0%
10. I would focus on the pre-specified analyses and trust the randomization.
Ignoring the potential for some residual confounding, I would trust the crude
risk difference and relative risk I calculated in question 2. There is the
potential that sex modifies the effects of idarubicin, but this study was not
powered sufficiently to explore that possibility and I would be wary to make
sex-specific recommendations based on this data.