Original Articles
Tao Zhou1, Zengfeng Ni2, Wenjing Fan1, Jing Xu3, Jing Qin1, Kaiyue Wang1, Rui Wang1 and Yaping Li1.
1 Department of Hematology, Xi'an International Medical Center Hospital, No. 777, Xitai Road, Gaoxin District, Xi'an 712000, Shaanxi Province, China.
2 Department of Hematology, The First Affiliated Hospital of Xi'an Medical University, No. 48, Fenghao West Road, Lianhu District, Xi'an 710000, Shaanxi Province, China.
3 Department of Hematology, Xi'an Gaoxin Hospital, No. 16, Tuanjie South Road, Gaoxin District, Xi'an City, Shaanxi Province, China.
Supplementary Figure 1. Three-knot restricted cubic spline analyses. Adjusted three-knot RCS associations of baseline platelet count with (A) severe bleeding (primary analysis, n=231, excluding patients with bleeding at diagnosis) and (B) 24-month OS (n=356). For severe bleeding, the overall association was not statistically significant (P=0.218) with no evidence of non-linearity (P=0.130). For 24-month OS, the overall association was statistically significant (P<0.001) and approximately linear (non-linearity P=0.069). Shaded areas show 95% CIs. The severe-bleeding RCS was adjusted for DIC and induction treatment class; the OS RCS was adjusted for age, WBC, bone marrow blasts, cytogenetic risk, induction class, ECOG, and AML type.
Supplementary Figure 2. Martingale-residual diagnostics for WBC functional form. Martingale residuals from the 24-month Cox model excluding WBC plotted against (A) WBC on the original scale and (B) log(WBC+0.1). Red curves represent loess smoothers; dashed horizontal lines indicate zero. Mild curvature was observed at higher WBC values, but formal specification tests did not reject the linear form (likelihood-ratio P=0.0996; spline nonlinear-component P=0.0779).
Supplementary Table 1. Univariable and multivariable logistic regression analysis of severe bleeding (CTCAE grade ≥ 3) in the full cohort (sensitivity analysis, n=356, 61 events).
| Variable | Type | OR | 95% CI | P value |
|---|---|---|---|---|
| PLT group (Low vs. Normal/High) | Univariable | 3.044 | 1.555–5.962 | 0.001 |
| WBC, × 10⁹/L | Univariable | 1.017 | 1.011–1.024 | <0.001 |
| DIC: No (vs. Yes) | Univariable | 0.260 | 0.133–0.509 | <0.001 |
| HMA+VEN (vs. intensive) | Univariable | 2.401 | 1.258–4.582 | 0.008 |
| Other regimen (vs. intensive) | Univariable | 1.673 | 0.636–4.402 | 0.297 |
| ECOG ≥2 (vs. 0–1) | Univariable | 2.344 | 1.321–4.158 | 0.004 |
| Secondary/therapy-related AML | Univariable | 2.028 | 1.054–3.902 | 0.034 |
| PLT group (Low vs. Normal/High) | Multivariable | 2.753 | 1.318–5.749 | 0.007 |
| WBC, × 10⁹/L | Multivariable | 1.018 | 1.010–1.025 | <0.001 |
| DIC: No (vs. Yes) | Multivariable | 0.335 | 0.156–0.721 | 0.005 |
| HMA+VEN (vs. intensive) | Multivariable | 2.875 | 1.370–6.034 | 0.005 |
| Other regimen (vs. intensive) | Multivariable | 2.729 | 0.926–8.042 | 0.069 |
| ECOG ≥2 (vs. 0–1) | Multivariable | 2.579 | 1.324–5.024 | 0.005 |
| Secondary/therapy-related AML | Multivariable | 2.065 | 0.961–4.433 | 0.063 |
This sensitivity analysis used the original severe-bleeding endpoint (all 356 patients, 61 events) and removed the baseline-bleeding covariate to avoid circularity with the outcome. The model additionally included ECOG performance status and AML type. Low baseline platelet count was associated with severe bleeding (adjusted OR=2.753, 95% CI 1.318–5.749, P=0.007). Abbreviations: OR, odds ratio; CI, confidence interval; DIC, disseminated intravascular coagulation; ECOG, Eastern Cooperative Oncology Group; HMA+VEN, hypomethylating agent plus venetoclax.
Supplementary Table 2. Comparison of baseline characteristics between patients censored before 24 months and those with adequate follow-up.
| Variable | Censored <24m (n=88) | Adequate follow-up (n=268) | P value | SMD |
|---|---|---|---|---|
| Age, years | 50.7±13.6 | 54.6±13.0 | 0.018 | 0.292 |
| WBC, × 10⁹/L | 61.8±40.6 | 57.2±39.9 | 0.347 | 0.115 |
| Hb, g/L | 78.2±18.7 | 78.2±19.9 | 0.998 | 0.000 |
| Baseline PLT, × 10⁹/L | 59.9±48.0 | 55.4±44.0 | 0.417 | 0.098 |
| BM blast, % | 58.8±16.7 | 60.0±17.5 | 0.585 | 0.068 |
| LDH, U/L | 519.8±229.3 | 513.5±220.2 | 0.820 | 0.028 |
| Fibrinogen, g/L | 2.6±0.9 | 2.4±0.9 | 0.059 | 0.236 |
| D-dimer, mg/L | 5.0±3.1 | 5.4±2.9 | 0.234 | 0.145 |
Early-censored patients were younger than those with adequate follow-up (50.7 vs. 54.6 years, P=0.018). No other baseline characteristics differed significantly (all P>0.05). Abbreviations: WBC, white blood cell count; Hb, hemoglobin; PLT, platelet count; LDH, lactate dehydrogenase; FIB, fibrinogen; DDimer, D-dimer; SMD, standardized mean difference.
Supplementary Table 3. Common-outcome scenario analyses for the association of low baseline platelet count with 24-month overall survival.
| Scenario | HR | 95% CI | P (Cox) | P (log-rank) | N | Events |
|---|---|---|---|---|---|---|
| Common-outcome: all-censored died | 2.012 | 1.472–2.750 | <0.001 | <0.001 | 356 | 194 |
| Common-outcome: all-censored survived | 2.276 | 1.469–3.526 | <0.001 | <0.001 | 356 | 106 |
Common-outcome scenario: all-censored died assumed all 88 patients censored before 24 months died within 24 months (HR=2.012, 95% CI 1.472–2.750, P<0.001). Common-outcome scenario: all-censored survived assumed they all survived to 24 months (HR=2.276, 95% CI 1.469–3.526, P<0.001). Both common-outcome scenarios produced hazard ratios in the same direction as the primary analysis; however, these scenarios do not address differential informative censoring between platelet groups.
Supplementary Table 4. Sensitivity analysis restricted to patients receiving intensive induction (DA or IA regimens, n=264).
| Analysis | OR/HR | 95% CI | P value | N | Events |
|---|---|---|---|---|---|
| Severe bleeding (primary endpoint) | 1.821 | 0.462–10.038 | 0.407 | 174 | 9 |
| 24-month OS | 2.402 | 1.273–4.532 | 0.007 | 264 | 67 |
In patients receiving intensive induction, low baseline platelet count was associated with 24-month overall survival (adjusted HR=2.402, 95% CI 1.273–4.532, P=0.007; 67 deaths). The severe-bleeding association was not statistically significant in this subgroup (OR=1.821, 95% CI 0.462–10.038, P=0.407; n=174, 9 events). Abbreviations: HR, hazard ratio; OR, odds ratio; CI, confidence interval.
Supplementary Table 5. Center and calendar-period sensitivity analyses.
| Analysis | Effect measure | OR/HR | 95% CI | P value | N | Events |
|---|---|---|---|---|---|---|
| Center-adjusted 24-month OS | HR | 1.896 | 1.192–3.017 | 0.007 | 356 | 106 |
| Center × PLT-group interaction for 24-month OS | Interaction P | — | — | 0.186 | 356 | 106 |
| 2020–2022: 24-month OS | HR | 1.614 | 0.975–2.672 | 0.063 | 279 | 88 |
| 2023–2025: 24-month OS | HR | 3.048 | 0.880–10.560 | 0.079 | 77 | 18 |
| 2020–2022: severe bleeding | OR | 1.253 | 0.413–4.178 | 0.696 | 181 | 16 |
| 2023–2025: severe bleeding | OR | 4.994 | 0.419–1385.252 | 0.224 | 50 | 4 |
The center-adjusted model added participating center to the prespecified 24-month Cox model. The interaction P value tests whether the platelet-group association varied by center; a non-significant result does not prove homogeneity. Calendar-period analyses were exploratory. Estimates for severe bleeding, particularly in 2023–2025 (4 events), were highly imprecise and should not be interpreted as evidence of no association. OR, odds ratio; HR, hazard ratio; CI, confidence interval; OS, overall survival; PLT, platelet count.