1.Shown below are the 30 survival times (months) of 30 melanoma patients, 11 of
1.Shown below are the 30 survival times (months) of 30 melanoma patients, 11 of which were treated with the
immunotherapy BCG and 19 treated with the immunotherapy c. parvum. The “+” indicates
censoring.
BCG: 3.9 5.4 7.9 10.5 16.6+ 16.9+ 17.1+ 19.5 23.8+ 33.7+
33.7+
c parvum: 6.9 7.7 7.8+ 8.0 8.2+ 8.2+ 8.3 10.8+ 11.0+ 12.2+
12.5+ 14.8+ 16.0+ 18.1+ 21.4+ 23.0+ 24.4 24.8+ 26.9+
a. Estimate the survival curves for each treatment using the Kaplan-Meier method by hand.
b. Plot the curves on the same graph.
2. As one example of survival data, we looked at the remission duration times on 42 patients with acute
leukemia, half of which received 6-MP and the other half a placebo. The dataset LEUKEMIA.DAT on
Canvas has with the following format:
Variable Columns
Group (1 = 6-MP, 2 = Placebo) 1
Remission time (in weeks) 3-4
Censoring status (1=censored, 2=not censored) 6
a. Estimate the Kaplan-Meier remission duration curves using SAS.
b. Test to determine if the groups are significantly different and interpret your results.
3. A study examines the effect of stopping smoking on survival in advanced lung cancer patients. Of 137
patients, 122 died by the end of the study period. Survival time is taken from date of diagnosis to death or
last contact. The following variables are included in the analysis:
Variable Coding _
1. Smoking status 0: continues smoking after Dx
1: stopped >1 yr prior to Dx
2. Performance status 0 : ≥ 80
1 : ≤ 70
3 & 4. Weight loss prior to therapy (3 categories)
(2 dummy variables created: WTL1 and WTL2)
None WTL1=0 and WTL2=0
≤ 10% WTL1=1 and WTL2=0
> 10% WTL1=0 and WTL2=1
5 & 6. Histology (3 categories)
(2 dummy variables created: HIST1 and HIST2)
Squamous Cell HIST1=0 and HIST2=0
Adenocarcinoma HIST1=1 and HIST2=0
Larqe cell HIST1=0 and HIST2=1
The table below summarizes the results of fitting several proportional hazard models. Regression coefficients &
log likelihoods are shown.
Model
Variables 1 * 2 3 4 5
Smoking(SMK) -0.552 -0.162 -0.032
Perf. Status (PS) 0.561 0.588 0.513
WTL1 0.056 0.058 0.063
WTL2 0.083 0.094 0.110
HIST1 0.061 0.066 0.132
HIST2 0.811 0.766 0.435
SMKxPS interaction -0.006
SMKxWTL1 ” 0.012
SMKxWTL2 ” 0.003
SMKxHIST1 ” -0.002
SMKxHIST2 ” -0.105
Log likelihood -486.54 -483.20 -476.12 -475.84 -472.73
* No variables in model
a. Compare the unadjusted survival curves for smoking status. If significant, describe the relationship
(direction, magnitude).
b. Determine whether the interaction terms of smoking status with performance status, weight loss and
histology taken together affect survival.
c. Assuming no interactions, compare the survival curves for smoking status adjusted for performance
status, weight loss and histology. If significant, describe the relationship.
d. What are your conclusions?
Leave a Reply