Group 4 HW 1 - Report

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Effect of Indinavir plus Zidovudine and

Lamivudine on Preventing AIDS


Progression among HIV-infected
Patients:
A Randomized Controlled Trial

Homework 1: Survival Data Analysis (0383)


2022-2023

2nd year Master of Statistics and Data Science


Hasselt University

Group members:
Melvin Estolano (2159122)
Quynh Long Khuong (2159280)
Jessa Mae Lastimoso (2159120)
Jose Carlos Cortiñas Porras (2055631)

Submission Date: October 19, 2022

Lecturer:
Prof. Dr. Tomasz Burzykowski
Group 4: Survival Data Analysis Homework 1

1 Introduction
Human Immunodeficiency Virus (HIV) is a virus that attacks the body’s immune system. If HIV is not
treated, it can lead to Acquired Immunodeficiency Syndrome (AIDS) which is a disease that progresses
to immune system failure, this leads to the development of other infections and potentially deadly
cancers. HIV is spread mainly by unprotected sex, contaminated blood transfusions, hypodermic
needles, and from mother to child. While there is no available cure, antiretroviral (ARV) treatment
can slow the course of the disease leading to a near-normal life expectancy.
A clinical trial of HIV-infected patients was conducted to study the effect of adding Indinavir(IDV) to
zidovudine and lamivudine(ZDV+LAM) treatment on time AIDS-free survival time while taking into
account other variables that might have an effect.

2 Methodology
2.1 Study setting and participants
This is a randomized controlled trial included 1148 patients with HIV. Patients were recruited in the
study if they (1) had no more 200 baseline TCD4 cells per cubic millimeter, and (2) were treated
with ZDV at least 3 months prior to the study. Patients were randomly assigned to the control group
(ZDV+LAM) or experimental group (ZDV+LAM+IDV).

2.2 Outcome
The primary outcome of this study was AIDS-free survival time. Patients were followed until having
AIDS-defining diagnosis, death, loss to follow-up, or end of the study, whichever occurs first.

2.3 Statistical analysis


The survival probabilities were calculated using the Kaplan-Meier method. The standard errors (SE)
were estimated based on the Greenwood’s formula, and the corresponding point-wise confidence in-
tervals were derived using the log-log transformation that keeps the values within 0 and 1.
The log-rank tests were utilized to compare the survival curves of the two treatments and other
variables including sex, race, history of IV drug use, and hemophiliac status. Age was divided into
three groups (< 35, 35 − 49, and 50+) [1] while the duration of ZDV use was classified into five groups
(< 1y, 1− < 2y, 2− < 4y, 4− < 6y, ≥ 6y). As age, Karnofsky performance scale score, and ZDV use
duration are ordinal variables, the log-rank test for trend was applied.
The stratified log-rank test was used to see the comparison of the survival curves across the Karnofsky
performance scale which is an assessment of the progression of an illness [2] and also for the variable
age. All the analyses were conducted in R version 4.2.1 [3] with packages survival and survminer,
and SAS using PROC LIFETEST procedure at a significance level of 5%.

2.4 Data Description


Seven variables were included in the study which are treatment, age, sex, race, IV drug use, Hemophilia
diagnosis, duration of ZDV use, and Karnofsky Performance Scale. Upon data exploration, it was also
seen that 82 patients have CD4 cells greater than 200 but were still included in the clinical trial. Two

1
Group 4: Survival Data Analysis Homework 1

datasets were used in the analysis, (1) using the original dataset and (2) a dataset that contain only
patients who met the eligibility criteria (i.e., TCD4 ≤ 200 cell per cubic millimeter at baseline).

3 Results and Discussion


Figures 1 show the Kaplan-Meier survival curve estimates for the different variables. Based on the
visual inspection of the graphs, first, we observed that across all variables, the median time to event
was not reached. Secondly, we see that the estimated survival curves for those participants in the
experimental group were greater as compared with the control group. Regarding other variables, we
see that for sex, race, history of IV drug use, hemophilia, and duration of ZDV use, the estimated
survival curves were similar across levels. Age group and Karnofsky Performance Scale score variables
had the most difference across the levels’ estimated survival curve.

Table 1: Log-rank test and Log-rank test for trend for independent variables

Treatment Agea Sex Race IV drug Hemo Karnofa ZDVa


Test Statistic 10.5 13.22 0.077 5.7 3.2 0 -47.96 2.97
Original dataset
p-value 0.001 0.036 0.781 0.223 0.202 1 <0.001 0.812
Test Statistic 11.4 12.81 0.088 5.4 2.8 0 -46.77 5.77
Reduced dataset
p-value <0.001 0.039 0.767 0.249 0.247 1 <0.001 0.641
a
Log-rank test for trend

Table 1 summarizes the results of the log-rank test. Comparisons involving nominal variables were
tested using a log-rank test, that is for: treatment, sex, race, IV drug, and hemophilia. On the other
hand, log-rank test for trend were applied for age group, Karnofsky performance scale score, and ZDV
class to reflect the ordinal nature of these variables. In both datasets, the differences in the estimated
survival curves across the levels of treatment, age group, and Karnofsky performance scale score were
statistically significant at 5% significance level. Meaning, for treatment variable, the experimental
group (ZDV+LAM+IDV) increases the probability of being AIDS-free as compared to the control
group (ZDV+LAM). For age group and Karnofsky performance scale score, the result indicates the
estimated survival curves are ordered corresponding to the order of the groups.

Table 2: Log-rank test for treatment stratified by other variables

Age Karnof
χ2df 10.8(1) 11(1)
Original dataset
p-value 0.001 <0.001
χ2df 11.7(1) 11.6(1)
Reduced dataset
p-value <0.001 <0.001

Table 2 shows the log-rank test results for treatment survival curves stratified on the covariates in the
dataset. Results showed consistency in both datasets. That is, after allowing for the stratification
for both age group and Karnofsky performance scale score, the estimated survival curve between the
levels of the treatment were significantly different at 5% level of significance.

2
Group 4: Survival Data Analysis Homework 1

Figure 1: Kaplan-Meier Survival Curve Estimates

4 Conclusion
The randomized clinical trial was done to see the effect of the new treatment (ZDV+LAM+IDV)
on the AIDS-free survival time. The results showed that combination of (ZDV+LAM+IDV) is more
effective in preventing AIDS progression as compared to the (ZDV+LAM) treatment. Moreover,
the progression of AIDS is slower among younger patients and patients who have higher Karnofsky
performance scale score. Lastly, the effect of IDV treatment is consistent after taking into account
grouping variables age group and Karnofsky performance scale score.

3
Group 4: Survival Data Analysis Homework 1

5 Appendix
R Codes
# Estimation the Kaplan - Meier table
aids _ KM _ tr <- survfit ( Surv ( time , censor ) ~ Treatment , data = df ,
type = c ( " kaplan - meier " ) ,
error = c ( " greenwood " ) ,
conf . type = c ( " log - log " ) )
summary ( aids _ KM _ tr )
# Ploting of survival curve
ggsurvplot ( aids _ KM _ tr , risk . table = TRUE , censor . shape = " '" ,
legend . labs = c ( " Control " , " Experimental " ) ,
risk . table . col = " strata " , legend . title = " " ,
palette = " Dark2 " , risk . table . y . text = FALSE ,
ylab = " AIDS - free ␣ probability " , xlab = " Day " )
# Log - rank test
survdiff ( Surv ( time , censor ) ~ Treatment , data = df )
# Stratified log - rank test
survdiff ( Surv ( time , censor ) ~ Treatment + strata ( karnof ) , data = df )
SAS Codes

/*Log-rank Trend Tests*/


proc lifetest data=aids;
time time*censor(0);
strata karnof_gr / group=idv test=(log-rank); run;

References
[1] Peter Ssebutinde, Imelda T Kyamwanga, Eleanor Turyakira, Stephen Asiimwe, and Francis Baju-
nirwe. Effect of age at initiation of antiretroviral therapy on treatment outcomes; a retrospective
cohort study at a large hiv clinic in southwestern uganda. PLoS One, 123(3):958–965, 2013.

[2] Buck Christenssen. Karnofsky performance status scale. https://emedicine.medscape.com/


article/2172510-overview#a1, 2018. Accessed:17/10/2022.

[3] R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for
Statistical Computing, Vienna, Austria, 2022.

You might also like