Survival Analysis in Stata: stset and sts – Kaplan-Meier

1. Declaring Survival Data with stset

Purpose:The stset command in Stata designates your dataset as survival (time-to-event) data. This step is essential before performing any survival or time-to-event analysis.

Core Syntax:

stset timevar, failure(eventvar)

• timevar: The variable representing follow-up time (e.g., number of days until event or censoring).
• failure(eventvar): Indicator variable for the event (1 = event occurred, 0 = censored).

Implanon Example:

stset day, failure(remove)

• day: Time from Implanon insertion to removal or last follow-up (censoring).
• remove: 1 if Implanon was removed, 0 if still in place at last follow-up.

What Stata Reports:

• Number of records, number of events
• Total person-time observed
• Range of observed time

2. Key stset Options in Clinical Datasets

Scenario:If you want to measure analysis time from date of birth (dob), but subjects only become at risk when they enroll (enrollment), use:

stset day, failure(remove) origin(time dob) enter(time enrollment)

This setup measures time from birth but only counts from when the participant actually joined the study.

3. Kaplan-Meier and Survival Analysis with sts

After using stset, you can perform nonparametric survival analyses and visualize results using the sts suite of commands.

Typical Tasks:

• Plot the survivor function (Kaplan-Meier curve):sts graph
• Plot cumulative incidence (1 – survival):sts graph, failure
• Display life-table details:sts list
• Visualize survival by subgroups (e.g., living with husband):sts graph, by(notlivingtogether)
• Add confidence intervals:sts graph, by(notlivingtogether) ci
• Compare groups using the log-rank test:sts test notlivingtogether

Example:

sts graph, by(notlivingtogether) ci

This command will display Kaplan-Meier curves for women living with vs not living with their husband, with confidence intervals.

4. Sample Analysis Workflow

// 1. Set up survival data structure
stset day, failure(remove)

// 2. Plot the hazard function (instantaneous event rate)
sts graph, hazard

// 3. Plot cumulative incidence (probability Implanon was removed)
sts graph, failure

// 4. List survival probability at specific times, by living arrangement
sts list, surv at(0 90 180 365) by(notlivingtogether)

5. Interpretation Tips

• Kaplan-Meier curves: A sharper decline indicates a higher rate of the event.
• sts list output: Look at the columns—number at risk, number of events, number lost (censored), survivor function, standard error, and confidence intervals.
• Group comparisons: Use by() to examine differences in survival between groups (e.g., exposure vs control).

6. Handling Censoring and Multiple Records

If each subject can have multiple records (such as with time-dependent covariates), always use the id() option to uniquely identify subjects:

stset day, id(woman_id) failure(remove)

This tells Stata which rows belong to the same individual.

7. Essentials for Practice and Exams

• Begin with stset to structure your data.
• Always specify time, event, and subject ID (if applicable).
• Plot and tabulate survival and failure functions with sts.
• Use the log-rank test (sts test) to statistically compare survival between groups.
• Incorporate origin(), enter(), or scale() for complex entry or time-scale scenarios.