Prism's Kaplan-Meier survival analysis can now work directly with Multiple Variables data tables. This means you can keep your survival data in the standard database format used by most statistical software and clinical trial databases, rather than reformatting it into Prism's traditional Survival table format.

The survival analysis itself is identical: it uses the same calculations, generates the same plots, and performs the same log-rank tests. The only difference is how you organize your data and tell Prism which variables contain your time values, event indicators, and group assignments.

What's New

When you run survival analysis from a Multiple Variables table, you'll see a new Data tab in the analysis parameters dialog. This is where you assign variables to their roles:

•Status (censoring) variables: The variable that tells Prism whether each subject experienced the event you're tracking or was censored

•Time variable: The time values (days, weeks, months, etc.)

•Grouping variables: Optional categorical variables to generate multiple separate survival curves and to compare survival curves between groups

Everything else - the Method tab, Output tab, survival curves, log-rank tests - works exactly as it does with traditional Survival tables.

When to Use Multiple Variables Tables for Survival Analysis

Multiple Variables tables work particularly well for survival data because clinical and experimental databases almost always store this kind of data in long format (one row per subject with columns for survival time, event status, and grouping factors). If your data is already in this format, you can import it directly into Prism without restructuring.

Consider using Multiple Variables format when:

•You're importing data from clinical trial databases or other data management systems

•Your data includes multiple grouping factors you want to explore (treatment, stage, age group, etc.)

•You want to run multiple different survival analyses on the same dataset with different grouping variables

•You're working with data that's already in standard "tidy" format

•You need to perform other analyses on the same data (Cox regression, other statistical tests)

The traditional Survival table format still works great for simple survival studies where you're just comparing two or three groups and entering data directly into Prism.

How Survival Data is Structured

Survival analysis needs three key pieces of information for each subject:

1.Time value: How long was this subject followed? (days, months, years, etc.)

2.Event status: Did the event of interest happen, or was this observation censored?

3.Group membership (optional): Which treatment group, risk category, or other grouping does this subject belong to?

In a Multiple Variables table, each row represents one subject, and columns contain these different pieces of information.

Basic Survival Data Layout

Here's a typical survival dataset structure:

In this example, Patient 003 and Patient 006 were censored (Event = 0). This means that they were still alive at 365 days when follow-up ended, or perhaps they dropped out of the study after being observed on day 365. The others experienced the event (Event = 1) at the times shown.

When you run the analysis, you would assign:

•Status (censoring) variable: Event (the censoring indicator)

•Time variable: Survival_Days (the time values)

•Grouping variable: Treatment (to compare survival curves between groups)

Working with Multiple Status Variables

One powerful feature of using Multiple Variables tables is the ability to analyze multiple outcomes from the same dataset. For example, you might have variables for different types of events - death, disease progression, treatment discontinuation - and you want to create separate survival curves for each outcome.

You can assign multiple status variables (event indicators) in a single analysis. Prism will create separate survival curves for each response variable. If you also assign grouping variables, you'll get survival curves for every combination of response variable and group.

Example with multiple outcomes:

If you assign both Death and Progression as status variables, with Treatment as a grouping variable, Prism will create four survival curves:

•Death - Control

•Death - Drug_A

•Progression - Control

•Progression - Drug_A

This lets you easily explore different endpoints from a single analysis setup.

Working with Multiple Grouping Variables

Clinical trials and observational studies often involve multiple factors you might want to use for stratification - treatment group, disease stage, age category, performance status, and so on. With Multiple Variables tables, you can assign multiple grouping variables, and Prism will create survival curves for every unique combination.

Example with multiple grouping factors:

Assigning both Treatment and Stage as grouping variables would create four survival curves:

•Control-Early

•Control-Advanced

•Drug_A-Early

•Drug_A-Advanced

This makes it easy to see whether treatment effects differ by disease stage.

Keep in mind that more grouping variables means more curves and more statistical comparisons. With 2 treatments, 2 stages, and 2 age groups, you'd have 8 survival curves. Make sure you have enough events in each combination to produce meaningful results.

Status Variable: The Censoring Indicator

The status variable in survival analysis is your censoring indicator. This is the column that tells Prism whether each subject experienced the event or was censored. Prism needs to know which code you're using:

•1 = Event, 0 = Censored (most common convention)

•0 = Event, 1 = Censored (sometimes used in older datasets)

You'll specify which convention you're using on the Method tab of the analysis parameters dialog, just like you do with traditional Survival tables. Prism will then interpret your data correctly.

You may also choose to assign a categorical variable here. In this case, you may choose which level of your chosen categorical variable corresponds to a censored observation and which corresponds to an event. This could be something as straightforward as "Censored" and "Event" as the levels of the categorical variable, or using any other experiment-specific indicator that makes the most sense to you.