Reduced-intensity versus intensive induction in high-risk acute promyelocytic leukemia: a systematic review and meta-analysis

Highlights

• Reduced-intensity all-trans retinoic acid plus arsenic trioxide (ATRA+ATO) achieves equivalent complete remission (CR; 92%) and lower early mortality (9%) compared with intensive chemotherapy in high-risk acute promyelocytic leukemia (APL).
• Randomized controlled trial (RCT) and risk-of-bias sensitivity analyses confirm robustness, supporting reduced intensity as a first-line APL induction.

Introduction

Acute promyelocytic leukemia (APL) treated with all-trans retinoic acid (ATRA) and arsenic trioxide (ATO) achieves cure rates exceeding 90% in standard-risk patients (1). The Sanz risk classification defines high-risk disease as white blood cell (WBC) count greater than 10 × 10⁹/L at diagnosis (2). High-risk APL comprises 20%–30% of cases and carries early mortality (EM) rates of 15%–25% compared with 5%–10% in standard-risk patients (3–6).

The landmark APL0406 and AML17 trials established the superiority of ATRA+ATO in standard/intermediate-risk APL (1, 7). However, these trials excluded high-risk patients or included insufficient numbers for separate analysis, leaving optimal treatment for this population uncertain. Concerns about inadequate cytoreduction led many centers to continue intensive anthracycline-based induction for high-risk patients (8, 9).

Recent evidence has challenged this approach. Prospective studies have demonstrated excellent outcomes with ATRA+ATO combined with gemtuzumab ozogamicin (GO) for cytoreduction (10, 11). The APOLLO trial showed superior event-free survival (EFS) with reduced-intensity ATRA+ATO plus minimal idarubicin versus standard chemotherapy (12). The APL15 trial demonstrated that even chemotherapy-free ATRA+ATO with hydroxyurea achieves excellent outcomes with appropriate supportive care (13).

Despite this evidence, no pooled analysis has evaluated comparative outcomes between the reduced-intensity and intensive approaches in high-risk APL. We conducted this systematic review and meta-analysis to synthesize available evidence in this population.

Methods

This systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines (14) and is registered with PROSPERO (CRD420261293473).

We searched PubMed and the Cochrane Central Register of Controlled Trials from January 1, 2000, through December 31, 2025 (the full search strategy is provided in Search strategy in the Supplementary Material).

We included studies in English meeting the following criteria: adult patients with high-risk APL defined as WBC >10 × 10⁹/L at diagnosis (2); treatment with either the reduced-intensity or the intensive induction regimen as defined below; and reporting of at least one primary outcome. Primary outcomes were complete remission (CR) and EM (i.e., death within 30 days). Secondary outcomes included overall survival (OS), EFS, and relapse. Reduced-intensity regimens were defined as ATRA plus ATO-based induction with cytoreduction limited to one of the following: a) GO as a single dose of 6–9 mg/m²; b) idarubicin at one to two doses of 12 mg/m² each (cumulative ≤24 mg/m²) or daunorubicin at a cumulative dose ≤120 mg/m²; or c) hydroxyurea only or no cytotoxic agent (chemotherapy-free) (10–13). Intensive regimens were defined as ATRA-based induction with three or more anthracycline doses (cumulative idarubicin ≥36 mg/m² or daunorubicin equivalent ≥150 mg/m²). Mitoxantrone at the standard induction dose of 20 mg/m² per dose was classified as intensive regardless of the dose count. Detailed per-arm anthracycline dose data are provided in Supplementary Table 6. Studies were excluded if they enrolled exclusively pediatric patients (defined as patients aged <18 years), combined high-risk with standard-risk patients without reporting the outcomes separately by risk group, included fewer than 15 patients in the high-risk subgroup, or provided insufficient outcome data for extraction. The per-study age ranges and lower age bounds are summarized in Supplementary Table 1.

The treatment regimens were classified using a five-group schema based on the co-administered cytoreductive agents during induction, as follows: group 1—chemo-free (ATRA+ATO ± hydroxyurea); group 2A—ATRA+ATO+GO without anthracycline; group 2B—ATRA+ATO+low-dose anthracycline (one to two doses) ± GO; group 2C—other combination regimens including oral ATO or investigational agents; and group 3—intensive ATRA-based therapy with three or more anthracycline doses during induction. For studies reporting both an overall cohort arm and a dedicated high-risk subgroup arm, only the high-risk arm was used to avoid unit-of-analysis errors. Studies contributing two arms (e.g., standard-dose vs. low-dose anthracycline comparisons within ATRA+ATO) were treated as independent arms sharing a study-level identifier. For all primary comparative analyses, groups 1 through 2C were pooled as the reduced-intensity stratum, while group 3 constituted the intensive stratum.

Standard ATRA doses across the included studies ranged from 25 to 45 mg/m² per day (45 mg/m² per day in the majority of the Western studies and 25 mg/m² per day in Chinese studies). ATO was administered intravenously at 0.15–0.16 mg kg⁻¹ day⁻¹ across studies. GO was given as a single intravenous dose of 6 or 9 mg/m². Idarubicin, when used, was administered at 12 mg/m² per dose. Low-dose regimens included one to two induction doses, whereas intensive regimens included three or more doses. The individual study dosing details are provided in Table 1. Detailed study characteristics, including the consolidation and maintenance regimens, are provided in Supplementary Table 1.

Table 1. Characteristics of the included studies.

ATRA, all-trans retinoic acid; ATO, arsenic trioxide; GO, gemtuzumab ozogamicin; IDA, idarubicin; DNR, daunorubicin; AraC, cytarabine; HU, hydroxyurea; Mitox, mitoxantrone; std, standard dose; ld, low dose; RCT, randomized controlled trial.

We extracted the study characteristics (e.g., design, country, sample size, and follow-up), patient demographics, treatment protocols, and outcomes (i.e., CR, EM, OS, EFS, and relapse). For the time-to-event outcomes, we extracted hazard ratios (HRs) with 95% confidence intervals (95%CI). EFS was defined per primary study, and the definitions varied (Supplementary Table 5). The pooled EFS HR was restricted to the two RCTs with documented event definitions: AML17 (7) defined EFS as the time from randomization to relapse, death, or failure to achieve CR, while APOLLO (12) defined EFS as the time from randomization to relapse, death from any cause, or refractory disease.

Data were extracted independently by one reviewer (MK). EM was defined as death within 30 days of diagnosis, consistent with the majority of the included studies. CR was defined as morphological CR unless only molecular data were reported. In 23 of 25 arms, early deaths were included as treatment failures in the CR denominator. Relapse was recorded as morphological relapse unless only molecular data were available.

Risk of bias was assessed using the Cochrane RoB 2 tool for RCTs and ROBINS-I for observational studies (15, 16). RCTs were rated as low risk of bias, some concerns, or high risk of bias across five RoB 2 domains: randomization process (D1), deviations from the intended interventions (D2), missing outcome data (D3), measurement of the outcome (D4), and selection of the reported result (D5). Observational studies were rated across seven ROBINS-I domains: confounding (D1), selection of participants (D2), classification of interventions (D3), deviations from the intended interventions (D4), missing data (D5), measurement of outcomes (D6), and selection of reported results (D7).

Two prespecified sensitivity analyses examined the consistency of outcome ascertainment across the included studies. One study (23) applied a 60-day EM cutoff. The sensitivity analysis excluding this study’s intensive arm (n = 135) yielded pooled EM rates of 8.83% (reduced-intensity; unchanged) and 9.47% (intensive; 95%CI = 5.83%–15.02%), consistent with the primary estimates of 8.83% and 10.36%, respectively, supporting pooled inclusion. One arm (28) excluded eight patients who died within 3 days of diagnosis from the treatment denominator. The sensitivity analysis substituting an intent-to-treat denominator for this arm (16 CRs in 24 patients) yielded a pooled reduced-intensity CR rate of 90.59% (95%CI = 81.94%–95.33%), a 1.7 percentage point difference from the primary estimate, confirming robustness.

Random-effects meta-analysis was performed using R version 4.5.2. Proportions were pooled using generalized linear mixed models with logit transformation and inverse variance weighting. The HRs for time-to-event outcomes were pooled using the generic inverse variance method with random effects. Between-study variance (τ²) was estimated using maximum likelihood for proportions and restricted maximum likelihood for hazard ratios. CIs were calculated using the Hartung–Knapp–Sidik–Jonkman adjustment. Between-study heterogeneity was quantified using the I² statistic. For proportion analyses (i.e., CR, EM, and relapse), I² is reported separately within each intensity subgroup, whereas for the EFS HR analysis, I² reflects the single pooled estimate across contributing trials. Publication bias was assessed visually using funnel plots. Sensitivity analyses were performed by restricting to RCTs only, low risk-of-bias studies only, and leave-one-out analysis. The full sensitivity analysis results are presented in Supplementary Table 3. Subgroup analyses compared the outcomes between the reduced-intensity and intensive regimens. Exploratory WBC-stratified analysis examined CR by median WBC level (<30 vs. ≥30 × 10⁹/L). Statistical significance was defined as a two-sided p-value less than 0.05.

For treatment arms contributing zero events to a proportion outcome, a continuity correction of 0.5 was applied before logit transformation (Haldane–Anscombe correction). Between-study meta-regression examining the association of the WBC stratum (median WBC <30 vs. ≥30 × 10⁹/L) with the CR rate was performed using the method of moments estimator. Given the small number of contributing arms (k = 11), this analysis was considered exploratory and no formal test of residual heterogeneity was conducted. The certainty of evidence for each outcome was assessed using the GRADE framework and is summarized in Supplementary Table 4.

Results

The database search identified 325 unique records. After title and abstract screening, 43 full-text articles were assessed for eligibility. Of these, 17 studies met the inclusion criteria, providing 25 treatment arms with high-risk APL data totaling 1,323 patients: 11 reduced-intensity arms (449 patients, 33.9%) and 14 intensive arms (874 patients, 66.1%) (Figure 1).

The 17 included studies comprised four RCTs, four prospective cohort studies, and nine retrospective cohort studies (Table 1). Studies were published between 2006 and 2025, with patient enrollment spanning from 2002 to 2023. The median follow-up ranged from 16 months in the earliest study (18) to 102 months in the longest [23]. The reduced-intensity arms included four GO-based regimens, five arms using low-dose anthracycline with or without GO, and two chemotherapy-free ATRA+ATO arms (Table 1). All included arms employing ATO used intravenous administration during induction. Oral arsenic-based regimens were used as consolidation or maintenance in select studies (25), but not as the induction route captured in this analysis.

The risk of bias domain-level assessments are summarized in Figure 2 and Supplementary Table 2. All four RCTs were rated as having low risk of bias across all RoB 2 domains. Among the 13 observational studies assessed with ROBINS-I, 11 were rated moderate risk and two (26, 28) were rated serious risk, primarily due to serious confounding (D1) (26) and selection of participants concerns (D2) (28). Restriction of primary analyses to low-risk-of-bias RCTs only yielded CR rates of 92.1% (reduced-intensity) versus 90.5% (intensive, p = 0.70), consistent with the direction and magnitude of the full-evidence pooled estimates, supporting the robustness of the primary findings. Pooled estimates incorporating observational studies should be interpreted with caution given the predominance of moderate-risk studies and the potential for residual confounding.

CR data were available for 17 arms (840 patients): nine reduced-intensity arms (330 patients) and eight intensive arms (510 patients). The pooled CR rate was 92.3% (95%CI = 85.9%–95.9%) for the reduced-intensity versus 89.3% (95%CI = 82.6%–93.6%) for the intensive regimen (Figure 3A; Table 2). The 3.0 percentage point difference favoring the reduced-intensity regimen was not statistically significant (p = 0.35). Heterogeneity was low in the reduced-intensity group (I² = 0%) and moderate in the intensive group (I² = 56.1%), indicating consistent effect estimates in the reduced-intensity arms, but greater variability in the intensive therapy outcomes. An exploratory meta-regression of the CR proportion on arm-level median age (k = 14 arms) showed no significant association (β = −0.051 per year, p = 0.28) (Supplementary Figures 4, S5).

Table 2. Summary of outcomes by treatment intensity.

RCT, randomized controlled trial; HR, hazard ratio; NS, not significant.

EM data were available for 19 arms (936 patients): nine reduced-intensity arms and 10 intensive arms. The pooled EM was 8.8% (95%CI = 4.7%–15.9%) for the reduced-intensity versus 10.4% (95%CI = 6.8%–15.5%) for the intensive regimen (Figure 3B). The 1.6 percentage point absolute risk reduction with the reduced-intensity regimens was not statistically significant. Heterogeneity was moderate in both groups (I² = 36.7% for reduced-intensity and I² = 44.7% for intensive). Early deaths were primarily attributed to hemorrhage (intracranial or pulmonary), differentiation syndrome, and infection in both groups (3, 4, 29).

EFS HRs were available from two RCTs. The AML17 trial (57 patients) reported 5-year recurrence-free survival (RFS) of 100% versus 83% (HR = 0.12, 95%CI = 0.04–0.38, p = 0.03). The APOLLO trial (133 patients) reported 2-year EFS of 88% versus 71% (HR = 0.40, 95%CI = 0.17–0.93, p = 0.02). The pooled HR for EFS from the two RCTs was 0.23 (95%CI = 0.07–0.76, p = 0.015), representing a 77% reduction in EFS events favoring reduced-intensity regimens (Figure 4). Heterogeneity was moderate (I² = 63.8%), reflecting differences in the treatment protocols between trials.

Relapse data were available for 17 arms (836 patients). The pooled relapse rate was 3.9% (95%CI = 1.5%–9.6%) for the reduced-intensity versus 5.5% (95%CI = 2.7%–10.7%) for the intensive regimen (Supplementary Figure 1). Heterogeneity was low in both groups (I² = 7.4% and 22.2%, respectively).

Exploratory WBC-stratified analysis (11 arms, 569 patients) examined CR by median WBC category (Supplementary Figure 2). CR was 92.8% in the arms with median WBC <30 × 10⁹/L versus 88.7% at WBC ≥30 (p = 0.14). The reduced-intensity CR remained stable across strata (+0.9 points), while the intensive CR declined from 96.0% to 86.6% (−9.4 points) (Supplementary Figure 3), suggesting that the ATRA+ATO-based regimens maintained efficacy at higher WBC levels.

Sensitivity analyses confirmed the robustness of the primary findings (Supplementary Table 3). Restriction to low-risk-of-bias RCTs only (four arms from two trials with CR data) showed CR of 92.1% for the reduced-intensity versus 90.5% for the intensive regimen (p = 0.70). Leave-one-out analysis demonstrated that the pooled CR ranged narrowly from 90.2% to 91.5% regardless of which study was excluded, indicating that no single study disproportionately influenced the overall estimate.

Visual inspection of the funnel plot for CR (Figure 5) showed no obvious asymmetry suggesting publication bias.

The 5-year OS ranged from 70% to 100% in the reduced-intensity arms (pooled ~85%) and from 72.7% to 96% in the intensive arms. The high heterogeneity (I² = 57%–62%) likely reflects differences in the supportive care and consolidation strategies.

Discussion

This meta-analysis of 1,323 high-risk APL patients from 17 studies directly compares reduced-intensity versus intensive induction. ATRA+ATO-based regimens with cytoreduction achieved CR rates equivalent to that of intensive chemotherapy (92.3% vs. 89.3%, p = 0.35), with comparable EM (8.8% vs. 10.4%), significantly superior EFS (HR = 0.23, p = 0.015), and lower relapse rates (3.9% vs. 5.5%).

The AML17 trial randomized 57 high-risk patients to ATRA+ATO+GO versus ATRA+idarubicin (7). At 67 months follow-up, no patient relapsed in the reduced-intensity arm versus 19% in the chemotherapy arm (5-year RFS = 100% vs. 83%, HR = 0.12, p = 0.03). APOLLO randomized 133 high-risk patients to ATRA+ATO with limited idarubicin versus standard all-trans retinoic acid (ATRA) plus Idarubicin (AIDA) (12). The 2-year EFS was 88% vs. 71% (HR = 0.40, p = 0.02), with relapse rates of 1.8% vs. 17% (p = 0.008) and serious adverse events in 32% vs. 68% (p < 0.001). APL15 demonstrated that chemotherapy-free ATRA+ATO with hydroxyurea achieved 94% 2-year disease-free survival (DFS) in 21 high-risk patients (13). The MD Anderson GO-based regimen achieved 97.5% CR with 89.4% 5-year EFS at 62 months follow-up (10).

The WBC-stratified analysis showed that the reduced-intensity CR remained stable across strata (+0.9 points), while the intensive CR declined by 9.4 points at WBC ≥30, suggesting that ATRA+ATO-based regimens maintained efficacy at higher disease burden. These exploratory findings warrant prospective validation. A dedicated meta-analytic subgroup for WBC >50 × 10⁹/L was not feasible because no included arm had a median WBC ≥50 × 10⁹/L (maximum arm median, 40 × 10⁹/L). However, our WBC ≥30 stratum encompassed arms with WBC ranges extending to 339 × 10⁹/L (12) and 270 × 10⁹/L (22), confirming enrollment of patients with extreme leukocytosis. In this stratum, the reduced-intensity CR remained 92.5% versus 86.6% for the intensive regimens, representing the greatest divergence between approaches. Prospective studies with explicit WBC >50 × 10⁹/L stratification are needed to define outcomes in this highest-risk subgroup.

The mechanistic basis for non-inferior efficacy despite reduced chemotherapy intensity lies in the exquisite sensitivity of PML/RARA-positive promyelocytes to ATO-mediated degradation and ATRA-induced differentiation (1, 31). Unlike de novo AML, where the chemotherapy intensity correlates with disease control, APL blasts are uniquely vulnerable to targeted therapy regardless of presenting leukocytosis. This molecular susceptibility explains why reduced-intensity regimens maintain efficacy even at WBC ≥30 × 10⁹/L, where conventional chemotherapy has shown diminishing returns due to treatment-related complications (29).

Our pooled findings are consistent with real-world validation studies. The HARMONY registry (n = 1,438) demonstrated 7-year OS of 91% and relapse rates of only 3% with ATRA+ATO, with outcomes independent of the Sanz risk score (6). Similarly, the French multicenter cohort found 5-year OS of 89.9% versus 72.7% favoring ATRA+ATO over chemotherapy in high-risk patients (p = 0.035) (22). These real-world data, encompassing diverse practice settings and supportive care infrastructures, strengthen the generalizability of our meta-analytic conclusions beyond controlled trial populations.

For cytoreduction, GO (single dose of 6 or 9 mg/m²) provides targeted promyelocyte reduction without prolonged myelosuppression (30). For centers without GO access, low-dose idarubicin (one to two doses of 12 mg/m²) provides effective cytoreduction with substantially lower toxicity (12). Hydroxyurea represents an alternative cytoreductive strategy, particularly in resource-limited settings, and has demonstrated efficacy in the APL15 chemotherapy-free approach (13). The choice of agent should be individualized based on institutional access, patient comorbidities, and degree of leukocytosis, with all options demonstrating acceptable outcomes when combined with ATRA+ATO.

This analysis has limitations. Not all included studies reported all outcomes, resulting in different sample sizes for each analysis (840 for CR, 936 for EM, and 836 for relapse). The predominance of observational data introduces potential confounding by indication. However, the sensitivity analysis restricted to the two randomized trials showed consistent and even stronger results. Treatment heterogeneity exists within the reduced-intensity group, including the GO-based, low-dose anthracycline, and chemotherapy-free regimens. Geographic variations in supportive care practices may have affected the generalizability of the EM estimates (5, 6). The moderate heterogeneity observed in the EFS HR analysis (I² = 63.8%) reflects differences between the AML17 and APOLLO protocols.

Conclusions

This meta-analysis of 1,323 high-risk APL patients from 17 studies demonstrates that reduced-intensity ATRA plus ATO-based induction with cytoreduction achieves CR rates equivalent to that of intensive chemotherapy (92.3% vs. 89.3%, p = 0.35), with comparable EM (8.8% vs. 10.4%), significantly superior EFS (HR = 0.23, 95%CI = 0.07–0.76, p = 0.015), and lower relapse rates (3.9% vs. 5.5%). The convergent evidence from randomized trials and observational studies supports reduced-intensity ATRA+ATO-based regimens using GO or low-dose anthracycline as the preferred treatment approach for high-risk APL.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/frhem.2026.1851998/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Yingyu Chen, Fujian Medical University Union Hospital, China

Reviewed by: Harinder Gill, The University of Hong Kong, Hong Kong SAR, China

Yurday Öncül, Inonu Universitesi Tip Fakultesi, Türkiye