A Kaplan-Meier survival plot visualizes the probability of survival (or event-free time) over a time period using a step function. It is the standard method for estimating survival functions from time-to-event data, handling censored observations where the event has not yet occurred. The plot shows how survival probability decreases over time, with optional confidence intervals and comparison between groups.
""" anyplot.ai
survival-kaplan-meier: Kaplan-Meier Survival Plot
Library: pygal 3.1.0 | Python 3.13.13
Quality: 84/100 | Updated: 2026-05-11
"""
import os
import numpy as np
import pygal
from pygal.style import Style
# Theme tokens
THEME = os.getenv("ANYPLOT_THEME", "light")
PAGE_BG = "#FAF8F1" if THEME == "light" else "#1A1A17"
INK = "#1A1A17" if THEME == "light" else "#F0EFE8"
INK_MUTED = "#6B6A63" if THEME == "light" else "#A8A79F"
# Okabe-Ito palette (first series = brand green, second = vermillion)
BRAND = "#009E73"
SECONDARY = "#C475FD"
# Set seed for reproducibility
np.random.seed(42)
# Generate synthetic clinical trial survival data
# Two treatment groups: Standard (control) and Experimental (new treatment)
n_per_group = 80
# Standard treatment group - shorter median survival
survival_times_standard = np.concatenate(
[np.random.exponential(scale=18, size=50), np.random.exponential(scale=8, size=30)]
)
censored_standard = np.random.binomial(1, 0.25, size=n_per_group)
event_standard = 1 - censored_standard
# Experimental treatment group - longer median survival
survival_times_experimental = np.concatenate(
[np.random.exponential(scale=28, size=55), np.random.exponential(scale=12, size=25)]
)
censored_experimental = np.random.binomial(1, 0.30, size=n_per_group)
event_experimental = 1 - censored_experimental
# Calculate Kaplan-Meier for Standard treatment
order_std = np.argsort(survival_times_standard)
times_std_sorted = survival_times_standard[order_std]
events_std_sorted = event_standard[order_std]
censored_std_sorted = censored_standard[order_std]
unique_times_std = np.unique(times_std_sorted[events_std_sorted == 1])
survival_std = [1.0]
time_points_std = [0.0]
n_at_risk_std = len(times_std_sorted)
for t in unique_times_std:
d = np.sum((times_std_sorted == t) & (events_std_sorted == 1))
if len(time_points_std) > 1:
prev_t = time_points_std[-1]
lost = np.sum((times_std_sorted > prev_t) & (times_std_sorted < t))
n_at_risk_std -= lost
if n_at_risk_std > 0:
s = survival_std[-1] * (1 - d / n_at_risk_std)
else:
s = survival_std[-1]
time_points_std.append(t)
survival_std.append(s)
n_at_risk_std -= d
time_std = np.array(time_points_std)
surv_std = np.array(survival_std)
# Calculate Kaplan-Meier for Experimental treatment
order_exp = np.argsort(survival_times_experimental)
times_exp_sorted = survival_times_experimental[order_exp]
events_exp_sorted = event_experimental[order_exp]
censored_exp_sorted = censored_experimental[order_exp]
unique_times_exp = np.unique(times_exp_sorted[events_exp_sorted == 1])
survival_exp = [1.0]
time_points_exp = [0.0]
n_at_risk_exp = len(times_exp_sorted)
for t in unique_times_exp:
d = np.sum((times_exp_sorted == t) & (events_exp_sorted == 1))
if len(time_points_exp) > 1:
prev_t = time_points_exp[-1]
lost = np.sum((times_exp_sorted > prev_t) & (times_exp_sorted < t))
n_at_risk_exp -= lost
if n_at_risk_exp > 0:
s = survival_exp[-1] * (1 - d / n_at_risk_exp)
else:
s = survival_exp[-1]
time_points_exp.append(t)
survival_exp.append(s)
n_at_risk_exp -= d
time_exp = np.array(time_points_exp)
surv_exp = np.array(survival_exp)
# Create step function data for pygal XY chart
step_std = []
for i in range(len(time_std)):
if i > 0:
step_std.append((float(time_std[i]), float(surv_std[i - 1])))
step_std.append((float(time_std[i]), float(surv_std[i])))
step_exp = []
for i in range(len(time_exp)):
if i > 0:
step_exp.append((float(time_exp[i]), float(surv_exp[i - 1])))
step_exp.append((float(time_exp[i]), float(surv_exp[i])))
# Calculate 95% confidence intervals using Greenwood's formula
# Standard treatment CI
var_std = []
n_risk_std = len(times_std_sorted)
cumvar_std = 0.0
for i in range(len(time_std)):
if i == 0:
var_std.append(0.0)
else:
mask = times_std_sorted == time_std[i]
d = np.sum(mask & (events_std_sorted == 1))
if n_risk_std > 0 and d > 0:
cumvar_std += d / (n_risk_std * (n_risk_std - d + 0.001))
var_std.append(cumvar_std)
n_risk_std -= np.sum(mask)
se_std = surv_std * np.sqrt(var_std)
ci_upper_std = np.minimum(surv_std + 1.96 * se_std, 1.0)
ci_lower_std = np.maximum(surv_std - 1.96 * se_std, 0.0)
# Experimental treatment CI
var_exp = []
n_risk_exp = len(times_exp_sorted)
cumvar_exp = 0.0
for i in range(len(time_exp)):
if i == 0:
var_exp.append(0.0)
else:
mask = times_exp_sorted == time_exp[i]
d = np.sum(mask & (events_exp_sorted == 1))
if n_risk_exp > 0 and d > 0:
cumvar_exp += d / (n_risk_exp * (n_risk_exp - d + 0.001))
var_exp.append(cumvar_exp)
n_risk_exp -= np.sum(mask)
se_exp = surv_exp * np.sqrt(var_exp)
ci_upper_exp = np.minimum(surv_exp + 1.96 * se_exp, 1.0)
ci_lower_exp = np.maximum(surv_exp - 1.96 * se_exp, 0.0)
# Create CI band step data - upper bound then back via lower for fill effect
ci_band_std = []
for i in range(len(time_std)):
if i > 0:
ci_band_std.append((float(time_std[i]), float(ci_upper_std[i - 1])))
ci_band_std.append((float(time_std[i]), float(ci_upper_std[i])))
# Trace back along lower bound (reversed)
for i in range(len(time_std) - 1, -1, -1):
ci_band_std.append((float(time_std[i]), float(ci_lower_std[i])))
if i > 0:
ci_band_std.append((float(time_std[i]), float(ci_lower_std[i - 1])))
ci_band_exp = []
for i in range(len(time_exp)):
if i > 0:
ci_band_exp.append((float(time_exp[i]), float(ci_upper_exp[i - 1])))
ci_band_exp.append((float(time_exp[i]), float(ci_upper_exp[i])))
# Trace back along lower bound (reversed)
for i in range(len(time_exp) - 1, -1, -1):
ci_band_exp.append((float(time_exp[i]), float(ci_lower_exp[i])))
if i > 0:
ci_band_exp.append((float(time_exp[i]), float(ci_lower_exp[i - 1])))
# Get censored observation points - create vertical tick marks
# Each tick is a short vertical line segment at the survival probability
tick_height = 0.03 # Height of censored tick marks
censored_times_std = times_std_sorted[censored_std_sorted == 1]
censored_ticks_std = []
for ct in censored_times_std:
idx = np.searchsorted(time_std, ct, side="right") - 1
idx = max(0, min(idx, len(surv_std) - 1))
y_val = float(surv_std[idx])
# Each tick is a pair of points (vertical line segment)
censored_ticks_std.append((float(ct), y_val - tick_height / 2))
censored_ticks_std.append((float(ct), y_val + tick_height / 2))
censored_ticks_std.append((None, None)) # Break to create separate tick marks
censored_times_exp = times_exp_sorted[censored_exp_sorted == 1]
censored_ticks_exp = []
for ct in censored_times_exp:
idx = np.searchsorted(time_exp, ct, side="right") - 1
idx = max(0, min(idx, len(surv_exp) - 1))
y_val = float(surv_exp[idx])
censored_ticks_exp.append((float(ct), y_val - tick_height / 2))
censored_ticks_exp.append((float(ct), y_val + tick_height / 2))
censored_ticks_exp.append((None, None))
# Custom style with theme-adaptive colors
brand_ci_light = "rgba(0, 158, 115, 0.15)" # Brand green, light alpha
brand_ci_dark = "rgba(0, 158, 115, 0.15)"
secondary_ci_light = "rgba(196, 117, 253, 0.15)" # Secondary orange, light alpha
secondary_ci_dark = "rgba(196, 117, 253, 0.15)"
# CI band colors match data colors with transparency
ci_colors = (
brand_ci_light if THEME == "light" else brand_ci_dark,
secondary_ci_light if THEME == "light" else secondary_ci_dark,
BRAND,
SECONDARY,
BRAND,
SECONDARY,
)
custom_style = Style(
background=PAGE_BG,
plot_background=PAGE_BG,
foreground=INK,
foreground_strong=INK,
foreground_subtle=INK_MUTED,
colors=ci_colors,
title_font_size=28,
label_font_size=22,
major_label_font_size=18,
legend_font_size=16,
value_font_size=14,
stroke_width=3,
guide_stroke_dasharray="0",
major_guide_stroke_dasharray="0",
font_family="sans-serif",
)
# Create XY chart for survival curves
chart = pygal.XY(
width=4800,
height=2700,
style=custom_style,
title="survival-kaplan-meier · pygal · anyplot.ai",
x_title="Time (Months)",
y_title="Survival Probability",
show_dots=False,
stroke=True,
fill=False,
show_x_guides=False,
show_y_guides=False,
x_label_rotation=0,
truncate_legend=-1,
legend_at_bottom=True,
show_legend=True,
range=(0, 1.05),
include_x_axis=True,
margin=80,
spacing=50,
explicit_size=True,
)
# Add CI bands first (background, filled polygons for shaded effect)
# Don't show in legend to reduce clutter
chart.add("Standard 95% CI", ci_band_std, fill=True, stroke=False, show_dots=False, show_in_legend=False)
chart.add("Experimental 95% CI", ci_band_exp, fill=True, stroke=False, show_dots=False, show_in_legend=False)
# Add main survival curves on top (solid lines with higher stroke width)
chart.add("Standard Treatment", step_std, stroke_style={"width": 7}, fill=False)
chart.add("Experimental Treatment", step_exp, stroke_style={"width": 7}, fill=False)
# Add censored tick marks (vertical strokes on the curve) - don't show in legend
chart.add(
"Censored (Standard)",
censored_ticks_std,
stroke_style={"width": 4},
show_dots=False,
allow_interruptions=True,
show_in_legend=False,
)
chart.add(
"Censored (Experimental)",
censored_ticks_exp,
stroke_style={"width": 4},
show_dots=False,
allow_interruptions=True,
show_in_legend=False,
)
# Save outputs
chart.render_to_file(f"plot-{THEME}.html")
chart.render_to_png(f"plot-{THEME}.png")
Part of Kaplan-Meier Survival Plot on anyplot.ai.