cs.LGAug 23, 2026
SaveKMGen: A Skill-based Approach for Synthetic Individual Patient Data Generation
Organizations: Siebel School of Computing and Data Science University of Illinois Urbana-Champaign · Mayo Clinic Alix School of Medicine
Abstract
Individual patient data (IPD) from clinical trials is the substrate for survival modeling, meta-analysis, and safety research, yet IPD is rarely released. Prior work has addressed only half of this gap: reconstructing Kaplan-Meier (KM) curves from published plots -- typically requiring manual digitization or human-in-the-loop correction -- while offering no mechanism for generating the adverse-event (AE) streams that constitute the other half of a patient record. We introduce KMGen, the first end-to-end framework that (i) fully automates KM curve extraction at accuracy competitive with human-guided tools, and (ii) generates synthetic per-patient AE trajectories from public trial registry records. The extraction stage is a fully automated agentic pipeline -- an agent generates code to extract each step in the KM curve -- achieving a mean Integrated Absolute Error (IAE) of 0.0151 on a 32-plot benchmark spanning clean, edge-case, and adversarial conditions. The IPD generation stage decouples patient archetype extraction from statistical sampling: an LLM distills the trial record into arm-specific statistics, adverse events, patient demographics, and risk multipliers. A mechanistic sampler generates patient events via clinical archetypes, bootstrap rank-correlation coupling to the empirical KM curve (preserving the marginal survival distribution exactly), and cycle-based AE scheduling with an induction/maintenance split. Across three held-out oncology trials spanning an order of magnitude in cohort size and 30 independent regenerations per trial, KMGen achieves mean integrated KM absolute difference , sex/ECOG JSD on 5 of 6 demographic slots, and recovers of the top-15 AEs by exact MedDRA term under a single fixed parameter set. The pipeline is released as open source at https://github.com/chufangao/kmgen.
Figures & tables
Figure 1: KMGen pipeline. (a) Fully automated KM extraction (b) Individual Patient Data generation from extracted KM plot, ClinicalTrial.gov data, publication, and statistical analysis plan.
| Category | Mean IAE | Median IAE | Min | Max | Std | |
|---|---|---|---|---|---|---|
| Synthetic benchmark (ground truth: rendered step function) | ||||||
| Standard | 5 | 0.0051 | 0.0051 | 0.0033 | 0.0068 | 0.0015 |
| Edge | 7 | 0.0089 | 0.0070 | 0.0012 | 0.0237 | 0.0080 |
| Single stress | 12 | 0.0140 | 0.0087 | 0.0051 | 0.0540 | 0.0136 |
| Combo stress | 8 | 0.0283 | 0.0234 | 0.0012 | 0.0744 | 0.0231 |
| Synthetic (all) | 32 | 0.0151 | 0.0080 | 0.0012 | 0.0744 | 0.0164 |
Table 1: Extraction accuracy by plot category. The synthetic benchmark (top block, 32 plots) measures IAE against programmatically-rendered ground truth. The real-trial block compares the extracted published KM curve against the empirical KM curve reconstructed from sponsor-delivered IPD for three oncology trials (one arm per trial). IAE is the Integrated Absolute Error in ; lower is better.
Figure 2: Real (solid) vs. synthetic (dashed) KM curves. Mean on all trials; Mann–Whitney fails to reject the null on both smaller trials (mean ) but rejects on NCT00844649 in 29 of 30 seeds, driven by a 1.3-month median shift at .
| NCT03041311 | NCT02499770 | NCT00844649 | |
| Demographics (Stage B, mean std over 30 regens) | |||
| Sex JSD | 0.0054 0.0064 | 0.0033 0.0060 | 0.0005 0.0006 |
| ECOG JSD | 0.272 0.048 | 0.013 0.012 | 0.003 0.001 |
| Weight KS | 0.201 0.029 | 0.202 0.050 | 0.129 0.013 |
| Height KS | 0.217 0.043 | 0.166 0.041 | 0.102 0.014 |
Table 2: End-to-end fidelity and extraction quality. Synthetic cohorts at real enrollment. = lower is better, = higher is better. JSD , KS . Stage B values are reported as mean std across 30 independent regenerations of the synthetic cohort (seeds ); Stage A metrics are deterministic functions of the extracted configs and have no resampling variability.
Figure 3: Top-12 shared AE terms per trial, ranked by real-IPD prevalence. Top-15 overlap is – on the small trials and on NCT00844649 (means over 30 regenerations).
Figure 4: AE onset density. Synthetic onsets show cycle-length striping (a direct consequence of cycle-based sampling); real onsets are smoother due to continuous-time recording. Mean onset error is months.
Figure 5: Top-10 SOC distributions. The dominant real SOC is the top synthetic SOC in every trial, confirming correct organ-system targeting by the archetype risk multipliers.
Appendix figures & tables30 assets
Supplementary material from the paper’s appendix.
Appendix
| Factor | Key SOCs affected | Default | Evidence |
| Age 65 (“older”) | |||
| Blood & lymphatic | 1.5 | 21 ; 14 | |
| Cardiac | 1.8 | 25 | |
| Infections | 1.6 | 17 | |
| Nervous system | 1.3 | 1 | |
| Renal & urinary | 1.4 | 18 | |
Table 3: Default AE risk multipliers by patient factor and organ system. Per-SOC values are clinical estimates informed by the overall effect sizes in the cited studies (Appendix B ). All multipliers are relative to a baseline of 1.0.
Figure 6: Induction fraction sweep. The onset KS statistic is minimized at (KS ), with sharp degradation at low , validating the induction–maintenance phase split.
Figure 7: Onset KS as a function of induction fraction , showing a clear minimum near .
Figure 8: Patient propensity sweep. Burden KS decreases monotonically, confirming the lognormal frailty is necessary. The default is conservative; higher values further reduce the burden gap.
Figure 9: Burden KS as a function of patient propensity . KS decreases monotonically; the default is conservative — higher values further reduce the gap.
Figure 10: Archetype ablation. Removing risk multipliers preserves marginal survival but degrades AE timing and burden realism.
Figure 11: Metric sensitivity to synthetic cohort size. AE frequency metrics improve with sample size; KM fidelity and onset timing are stable, exhibiting bootstrap-level variance.
| OS KS | ||||||
|---|---|---|---|---|---|---|
| Trial (arm, ) | Extraction IAE | Extracted | Oracle | Extracted | Oracle | Propagated |
| NCT03041311 (Placebo, 53) | 0.0309 | 0.055 0.018 | 0.040 0.022 | 0.238 0.065 | 0.114 0.035 | 0.015 |
| NCT02499770 (E/P, 37) | 0.0345 | 0.056 0.018 | 0.049 0.018 | 0.156 0.035 | 0.147 0.035 | 0.007 |
| NCT00844649 (Gemcitabine, 430) | 0.0133 | 0.018 0.005 | 0.012 0.006 | 0.063 0.014 | 0.043 0.014 | 0.006 |
Table 4: Extraction-error propagation (EXP1). Extracted regenerates from the curve traced off the published figure; oracle regenerates from the KM reconstructed from held-out real IPD. Propagated is the difference in mean . Values are mean std over 30 regenerations. Unlike Table 2 , which pools all arms of a trial, these numbers are for the single matched arm, so the extracted-input column is not identical to the headline .
| Trial | Metric | ||||
|---|---|---|---|---|---|
| NCT03041311 | AEs/patient (real: 15.1) | 10.1 0.5 | 11.3 0.8 | 13.0 1.5 | 17.2 2.3 |
| Burden KS | 0.277 0.029 | 0.218 0.047 | 0.190 0.045 | 0.165 0.041 | |
| 0.049 | 0.049 | 0.049 | 0.049 | ||
| AE cosine | 0.852 | 0.850 | 0.852 | 0.861 | |
| NCT02499770 | AEs/patient (real: 12.8) | 6.9 0.6 | 7.7 0.6 | 9.5 1.2 | 10.5 1.2 |
| Burden KS | 0.457 0.057 | 0.407 0.044 | 0.344 0.057 | 0.352 0.057 |
Table 5: Propensity dispersion sweep (EXP2), 30 regenerations per cell. Burden approaches the real value monotonically in while is invariant (identical to four decimal places, as survival assignment does not depend on ) and AE frequency moves within noise. Best burden KS per trial in bold.
| NCT03041311 | NCT02499770 | NCT00844649 | |
| Demographics | |||
| Sex JSD | 0.0054 0.0064 | 0.0033 0.0060 | 0.0005 0.0006 |
| ECOG JSD | 0.272 0.048 | 0.013 0.012 | 0.003 0.001 |
| Weight KS | 0.204 0.039 | 0.200 0.056 | 0.125 0.017 |
| Height KS | 0.207 0.029 | 0.179 0.042 | 0.103 0.013 |
Table 6: Full Stage-B fidelity at the recommended (compare Table 2 , which uses ). Mean std over 30 independent regenerations, synthetic cohorts at real enrollment. Only the burden rows change materially; demographic, survival, and AE-frequency rows are within resampling noise of the values.
Figure 12: Weight, height, and BMI distributions (real vs. synthetic) for all three trials. KS statistics are in all cases; the truncated-Gaussian sampling reproduces the overall shape of real anthropometric distributions.
Figure 13: AE burden per patient. Dashed lines show means.
| # | Plot | Category | IAE | Med. AE | Med. OS err (mo) | Arms |
|---|---|---|---|---|---|---|
| 1 | stress_combo_flat_overlap | Combo stress | 0.0012 | 0.0005 | 0.000 | 2 |
| 2 | edge_high_survival | Edge | 0.0012 | 0.0007 | 0.000 | 2 |
| 3 | edge_near_flat | Edge | 0.0029 | 0.0016 | 0.032 | 2 |
| 4 | synthetic_002 | Standard | 0.0033 | 0.0016 | 0.095 | 2 |
| 5 | edge_ci_shading | Edge | 0.0035 | 0.0019 | 0.105 | 2 |
| 6 | synthetic_003 | Standard | 0.0040 | 0.0017 | 0.232 | 2 |
Table 7: Per-plot attempt-1 extraction metrics, sorted by IAE (best first).
Figure 14: synthetic_002 : Standard 2-arm plot over 36 mo. Clean color separation; steps tracked precisely. synthetic_003 : Standard 2-arm plot over 36 mo with tighter step spacing.
Figure 15: synthetic_001 : Two well-separated arms over 12 mo with clear step structure. Extraction markers match the steps visually with no systematic drift. synthetic_004 : Standard 2-arm, 18 mo. Accurate step tracking.
Figure 16: synthetic_005 : Standard 2-arm, 12 mo with steep drop. Vertical-drop detection picks up the final step.
Figure 17: edge_high_survival : Minimal events over 144 mo follow-up. Red and blue ring markers track both arms cleanly through the entire range, including closely-overlapping segments near 96–120 mo. edge_near_flat : Near-flat mortality profile. Ring density is high and follows the tight step structure closely.
Figure 18: edge_ci_shading : Confidence-interval shading suppressed by the color mask; curves tracked cleanly. edge_four_arms : 4-arm comparison; color separation and continuity tracking both hold.
Figure 19: edge_small_dense : Dense step structure on a 36-month scale; fine sampling required. edge_cumulative_incidence : Cumulative-incidence plot (curves rising). Monotonicity constraint inverted; rising structure recovered.
Figure 20: edge_multi_panel : Multi-panel layout (4 arms across subplots). Per-panel differs, inflating the aggregate IAE relative to what any single panel would show.
Figure 21: stress_gridlines : Light gridlines scrubbed by the 30% row/column coverage rule before tracing. stress_annotation_heavy : Multiple in-plot annotations masked before tracing.
Figure 22: stress_three_similar : Three similarly-hued arms. HSL mask separates them with a narrow margin. stress_multi_panel_clean : Clean multi-panel layout; per-panel causes minor ( ) mismatch between saved and recomputed IAE.
Figure 23: stress_jpeg_artifact : JPEG compression smears the edge; extraction stays within 1–2 px. stress_legend_overlap : Legend overlaps the plot area; masked as a rectangular exclusion zone before tracing.
Figure 24: stress_bw : Black & white with solid vs dashed distinction. Grayscale separation holds; dashed gaps bridged. stress_tiny : Low pixel density; small mask errors translate to larger absolute errors in .
Figure 25: stress_stretched_tall : Stretched vertically; tick calibration resolves the bbox but column density is lower. stress_blurry : Gaussian blur widens the curve edge symmetrically; centroid strategy absorbs it.
Figure 26: stress_stretched_wide : Stretched horizontally; abundant columns, shallow -axis resolution. stress_diep_like : Two-panel cumulative-incidence plot with 8 arms total and heavy overlap. Per-arm IAE ranges 0.042–0.072; color attribution is the bottleneck—several arms share hue bands, so the tracer occasionally jumps between arms in dense regions.
Figure 27: stress_combo_flat_overlap : Both arms stay between 0.95–1.00 across 60 mo. Step markers land on the curves throughout; the combination of overlap and flatness defeats itself—any pixel near the curve gives near-zero error. stress_combo_bw_overlap : B&W with overlapping solid and dashed curves; grayscale plus gap bridging resolves both.
Figure 28: stress_combo_4arm_tiny : 4 arms at tiny resolution; color separation holds but per-column sample count is low. stress_combo_tiny_bw : Tiny plus B&W; low pixel count compounds grayscale ambiguity.
Figure 29: stress_combo_tiny_blurry : Tiny plus blurry; edge widening dominates, yielding the highest IAE among these single-style combos. stress_combo_stretched_dense : Stretched aspect with dense step structure; occasional drift in densest regions.
Figure 30: stress_combo_jpeg_blurry_dark : Heavy JPEG artifacts and blur widen the curve edge. Centroid tracking keeps the extraction within a few pixels of truth, but the smeared edge produces noisier step timing. stress_combo_grid_annotation_bw : Adversarial combo: gridlines leak into the color mask, the “Median OS” diagonal leader line pulls the red markers off the true curve near , and B&W forces grayscale separation of solid vs dashed lines. Three failure modes stack.