TrialCurveATTRACTION-2 (2-year update): reconstructed Kaplan–Meier curves
A phase 3 study of nivolumab in previously treated advanced gastric or gastroesophageal junction cancer (ATTRACTION-2): 2-year update data. Curves rebuilt as patient-level data (Guyot 2012) and checked against the statistics printed on the figure.
| Publication | Gastric Cancer. 2019 Dec 20;23(3):510–9. doi: 10.1007/s10120-019-01034-7 |
|---|---|
| DOI | 10.1007/s10120-019-01034-7 |
| Registry | — |
| Phase | — |
| Conditions | — |
| Interventions | — |
| Sponsor | — |
| Enrolment | — |
| Source licence | Figure published under CC BY 4.0 |
Overall survival
From figure panel a · validated against the values printed on the figure · curve source: vision model
Reconstructed from the published curve. Median: Nivolumab 5.3 months; Placebo 4.1 months. Survival at 12 months: Nivolumab 27%; Placebo 12%. Hazard ratio Nivolumab vs Placebo 0.65 (95% CI 0.54–0.80).
| Arm | N | Events | Median (Months) |
|---|---|---|---|
| Nivolumab | 330 | 288 | 5.3 |
| Placebo | 163 | 152 | 4.1 |
- Nivolumab vs Placebo: reconstructed HR 0.65 (95% CI 0.54–0.80)
Landmark survival (reconstructed)
| Time | Nivolumab | Placebo |
|---|---|---|
| 6 mo | 46.2% | 34.7% |
| 12 mo | 27.5% | 11.8% |
| 18 mo | 16.9% | 3.9% |
| 24 mo | 10.8% | 3.3% |
| 36 mo | — | 1.6% |
Checks against the published figure
- Nivolumab median: printed 5.26, reconstructed 5.28 pass
- Placebo median: printed 4.14, reconstructed 4.09 pass
- hazard ratio (direction inferred): printed 0.62, reconstructed 0.653 pass
Progression-free survival
From figure panel b · validated against the values printed on the figure · curve source: vision model
Reconstructed from the published curve. Median: Nivolumab 1.6 months; Placebo 1.4 months. Survival at 12 months: Nivolumab 9%; Placebo 1%. Hazard ratio Nivolumab vs Placebo 0.61 (95% CI 0.50–0.75).
| Arm | N | Events | Median (Months) |
|---|---|---|---|
| Nivolumab | 330 | 277 | 1.6 |
| Placebo | 163 | 145 | 1.4 |
- Nivolumab vs Placebo: reconstructed HR 0.61 (95% CI 0.50–0.75)
Landmark survival (reconstructed)
| Time | Nivolumab | Placebo |
|---|---|---|
| 6 mo | 20.1% | 7.7% |
| 12 mo | 9.3% | 0.9% |
| 18 mo | 5.5% | 0.9% |
| 24 mo | 3.8% | — |
Checks against the published figure
- Nivolumab median: printed 1.61, reconstructed 1.58 pass
- Placebo median: printed 1.45, reconstructed 1.42 pass
- hazard ratio (direction inferred): printed 0.6, reconstructed 0.61 pass
Overall survival — CR+PR (best overall response)
From figure panel a · validated against the values printed on the figure · curve source: pixel measurement confirmed by vision model
Reconstructed from the published curve. Median: Nivolumab 26.6 months. Survival at 12 months: Nivolumab 87%.
| Arm | N | Events | Median (Months) |
|---|---|---|---|
| Nivolumab | 32 | 16 | 26.6 |
Landmark survival (reconstructed)
| Time | Nivolumab |
|---|---|
| 6 mo | 100.0% |
| 12 mo | 87.1% |
| 18 mo | 77.4% |
| 24 mo | 61.3% |
Checks against the published figure
- Nivolumab median: printed 26.6, reconstructed 26.6 pass
Overall survival — PD (best overall response)
From figure panel c · validated against the values printed on the figure · curve source: vision model
Reconstructed from the published curve. Median: Nivolumab 3.9 months; Placebo 3.7 months. Survival at 12 months: Nivolumab 11%; Placebo 8%. Hazard ratio Nivolumab vs Placebo 0.84 (95% CI 0.63–1.13).
| Arm | N | Events | Median (Months) |
|---|---|---|---|
| Nivolumab | 124 | 115 | 3.9 |
| Placebo | 79 | 75 | 3.7 |
- Nivolumab vs Placebo: reconstructed HR 0.84 (95% CI 0.63–1.13)
Landmark survival (reconstructed)
| Time | Nivolumab | Placebo |
|---|---|---|
| 6 mo | 25.4% | 29.8% |
| 12 mo | 11.3% | 7.5% |
| 18 mo | 6.6% | — |
| 24 mo | 4.7% | — |
Checks against the published figure
- Nivolumab median: printed 3.84, reconstructed 3.9 pass
- Placebo median: printed 3.75, reconstructed 3.72 pass
- hazard ratio (direction inferred): printed 0.83, reconstructed 0.841 pass
About these data
Pseudo-IPD reproduces the published curves and numbers at risk; the rows are not real patients. Cite the publication (DOI above) and TrialCurve. The source figure is © the authors (CC BY 4.0); the charts here are redrawn from the reconstruction.
Load these data in R or Python
# R
library(survival)
d <- read.csv("https://trialcurve.com/data/trials/attraction-2-2-year-update-1.csv")
survfit(Surv(time, event) ~ arm, data = d)
coxph(Surv(time, event) ~ arm, data = d)
# Python
import pandas as pd
from lifelines import KaplanMeierFitter
d = pd.read_csv("https://trialcurve.com/data/trials/attraction-2-2-year-update-1.csv")
for arm, g in d.groupby("arm"):
KaplanMeierFitter().fit(g.time, g.event, label=arm).plot_survival_function()
Suggested citation: Gastric Cancer. 2019 Dec 20;23(3):510–9. doi: 10.1007/s10120-019-01034-7 doi:10.1007/s10120-019-01034-7. Reconstructed data: TrialCurve (2026), https://trialcurve.com/trials/attraction-2-2-year-update
Report a problem with these data
Related trials
- ATTRACTION-2 (prior trastuzumab subgroup): Overall survival; Progression-free survival
- Preoperative chemoradiotherapy vs chemotherapy in gastric/GEJ cancer: Disease-free survival; Overall survival
- ATTRACTION-2 (Japanese subgroup): Overall survival; Progression-free survival
- JACOB (Chinese subgroup): Overall survival; Progression-free survival (investigator-assessed)
- GLOW: Overall survival; Progression-free survival; Progression-free survival (investigator-asses
- STARGATE: Overall survival; Progression-free survival (blinded central radiologic review)
Need reconstructed data for other trials? Upload any Kaplan–Meier figure, or ask about the TrialCurve oncology library.
Reconstruct a figure Browse all trials