Identification of favorable subgroups of pediatric high-hyperdiploid acute lymphoblastic leukemia—A retrospective multinational study

Abstract

B-cell acute lymphoblastic leukemia (B-ALL) with high-hyperdiploidy (HHD) represents the most prevalent genetic subtype of childhood ALL and is associated with favorable outcome measures.^1-3 HHD occurs in 25%–30% of B-ALL and is largely defined by a modal chromosome number (MCN) ranging from 51 up to 67.^{1, 2, 4} Although several study groups have proposed HHD-definitions to identify patients with particularly favorable outcome, there is no consensus regarding the use of HHD as a whole group or its subgroups for risk stratification. This issue is clinically relevant because HHD-ALL still accounts for a substantial proportion of relapses, but it also requires renewed consideration because modern treatment, including blinatumomab-based chemoimmunotherapy, is likely to change the prognostic contribution of genetic classifiers derived from chemotherapy-based cohorts.⁵

We therefore performed a retrospective multinational study within the international Berlin-Frankfurt-Munster (i-BFM) ALL Group to evaluate previously described HHD-subgroup definitions in European-based B-ALL trials. Our primary aim was not to compare the prognostic impact of HHD- versus non-HHD-B-ALL across trials. Rather, we aimed to determine, within each included study, which HHD-definition identified both the largest proportion of B-ALL cases and the most favorable outcome, followed by an assessment of consistency across studies. The study should thus be interpreted as a reference benchmark for ongoing and future HHD-ALL cohorts, rather than as a definitive guide for treatment de-intensification in contemporary protocols.

Data on basic characteristics, treatment, and outcome of non-Down syndrome B-ALL patients aged 1–18 years were collected from seven i-BFM ALL Group members. Overall, 11,530 B-ALL patients treated across nine multicenter trials were included (Supporting Information S3: Data Supplement), comprising 3231 patients with HHD- and 8299 with non-HHD-B-ALL.^6-15 Patients were treated in recruiting trials between 1998 and 2018. The median follow-up exceeded 5 years in all trials except ALL-IC BFM 2009.¹⁵ Apart from the EORTC-CLG 58951 trial, HHD was not used for risk stratification.⁶ For the COALL Study Group, the two consecutive trials COALL-07-03 and COALL-08-09 were analyzed together because they shared the same diagnostic work-up and central cytogenetic review and used closely related treatment backbones, whereas DCOG ALL10 and ALL11 were kept separate as they represented distinct, prospectively defined treatment-reduction strategies. This resulted in eight trial cohorts.^{8, 14}

For this study, HHD was defined as a chromosome number between 51 and 67.² Diagnostic bone marrow and/or peripheral blood samples were analyzed by standard G-banded karyotyping. Apart from the ALL-IC BFM 2009 and EORTC-CLG 58951 trials, all other trials had a central review of the karyotypes.^{6, 15} Patients diagnosed only by DNA-index, SNP-arrays, or other methodologies were excluded, as were patients with concomitant BCR::ABL1, ETV6::RUNX1, TCF3::PBX1, or KMT2A-involving fusions. Karyotypes were curated to exclude masked haploidy/hypodiploidy.¹⁶ Incomplete/substandard karyotypes were excluded, too. If only a range of MCN was available, the highest value was used for final categorization. Minimal residual disease (MRD) status at end of induction was available for the trial cohorts, but MRD methodology and thresholds were protocol-specific and not harmonized; therefore, MRD was not incorporated into the cross-trial subgroup analysis.

Eight previously described HHD subgroups were evaluated: triple trisomy of chromosomes 4, 10, and 17; double trisomy of chromosomes 4 and 10; trisomy 18; the UK good-risk (UK-GR) profile, defined by simultaneous gain of chromosomes 17 and 18, or gain of chromosome 17 or 18 in the absence of gain of chromosome 5 or 20⁷; and four MCN categories (51–53, 54–55, 56–67, and the composite 54–67).² These four MCN categories were pre-specified for the present study to capture chromosome-count thresholds previously associated with outcome, in particular the higher MCN range reported to be highly curable (e.g., 58–66 chromosomes in EORTC-CLG 58951), rather than being adopted from a single prior classification. Importantly, while the eight definitions are based on methodologies widely available, thus, making them applicable across different healthcare settings, the prognostic assessment of concurrent somatic mutations was not part of the study. Event-free survival (EFS), overall survival (OS), and cumulative incidence of relapse (CIR) were calculated at 5 years. Because treatment protocols, risk-stratification algorithms, and cytogenetic review procedures differed between trials, and centralized individual-patient data were not collected, comparisons across trials were considered descriptive. We did not perform pooled treatment-adjusted hazard-ratio analyses.

The frequency of HHD among B-ALL cases varied significantly, from 19% in ALL-IC BFM 2009 to 35% in EORTC-CLG 58 951 (P < 0.001, Supporting Information S1: Figure 1A).^{6, 15} Sex ratio, age, leukocyte count, and NCI-risk distribution among patients with HHD were broadly comparable across trials (Supporting Information S1: Figure 2), suggesting that part of the variability in HHD-frequency reflected differences in cytogenetic success rates and quality rather than major differences in case mix. As descriptive context, HHD-patients had lower 5-year CIR than other B-ALL patients in six of eight trials, but the difference was statistically significant in only three (Supporting Information S1: Figure 1B). Similarly, EFS and OS varied between protocols (Supporting Information S1: Figure 1C/D), emphasizing that HHD-outcome is influenced by the diagnostic and therapeutic setting.^6-15

The frequency of subgroups defined by specific chromosome gains varied between trial cohorts (Figure 1A). However, the pattern of variation was consistent for each subgroup (i.e., the same trial had the lowest frequency of each subgroup), suggesting that the differences were driven by the quality of the cytogenetic data (i.e., number of marker chromosomes) rather than by genuine biological heterogeneity. By contrast, MCN-defined subgroups were more comparable across datasets (Figure 1B) because they required only assessment of the total chromosome count. This observation might be relevant, because, while the MCN requires only assessment of the total chromosome count, the UK-GR profile is based on assessment of four specific chromosomes.⁷ Conversely, modern genome-wide methods may improve classification of both definitions and reduce the uncertainty inherent to conventional karyotyping.¹⁷

When the eight HHD-subgroup definitions were compared within each trial, MCN 54–67 and UK-GR most consistently combined broad representation with favorable outcome (Figure 2A–C). MCN 54–67 captured up to 28% of all B-ALL cases and UK-GR up to 26%. Five-year EFS ranged from 80.5% to 92.8% for MCN 54–67 and from 80.5% to 93.5% for UK-GR. Both components of MCN 54–67 contributed to this favorable profile: the MCN 54–55 (6%–11% of B-ALL) and MCN 56–67 (10%–17% of B-ALL) subgroups each showed broadly comparable 5-year EFS (82.5%–95.5% and 79.4%–92.2%, respectively; Supporting Information S2: Table 5), with neither subset alone clearly outperforming the composite. The lower end of these ranges, particularly in ALL-IC BFM 2009, should not be interpreted as uniformly excellent outcome; rather, it highlights protocol-level heterogeneity and argues against overgeneralization. In ALL-IC BFM 2009, no single definition clearly outperformed all others, supporting the need to test any proposed low-risk classifier prospectively within the respective treatment context.

To synthesize subgroup size and outcome without implying a formal pooled inferential model, we used a descriptive ranking system within each trial (Supporting Information S1: Figure 3A). Subgroup prevalence (Figure 1) and per-trial outcomes with 95% confidence intervals (Figure 2, Supporting Information S2: Table 5) are presented separately before this composite ranking, so that the contributions of coverage and prognosis can be appraised independently. This approach ranked each HHD-definition according to 5-year CIR, EFS, OS, and the proportion of total B-ALL and HHD-B-ALL cases captured. Across the eight trials, MCN 54–67 and UK-GR had the lowest cumulative scores (16 and 17, respectively), whereas MCN 51–53 had the worst possible cumulative score (64; Supporting Information S1: Figure 3B/C). This ranking deliberately combined outcome with subgroup size, because a clinically useful low-risk definition must identify a favorable yet sufficiently large and reliably ascertainable population; when outcome alone was considered, MCN 54–67 and UK-GR remained among the best and MCN 51–53 the worst. Because larger subgroups are rewarded, the composite MCN 54–67 scored better than either of its smaller components (MCN 54–55 and 56–67), reflecting its greater size and stability rather than a true outcome advantage over its parts. Patients with MCN 51–53 represented only 3%–7% of all B-ALL cases and had inferior EFS, ranging from 68.7% to 88.9%, indicating that this numerically related group should not be considered a favorable HHD-subset. Despite consistently higher 5-year CIR (9.0%–25.7%), 5-year OS in this subgroup generally remained relatively preserved (mostly >88%; Supporting Information S2: Table 5), indicating that the inferior EFS did not uniformly translate into excess mortality and that a proportion of relapses appeared salvageable.

The overlap between MCN 54–67 and UK-GR further supported their practical relevance (Supporting Information S2: Table 3). Because these two definitions ranked highest in our comparative analysis, we examined their concordance in detail; although the evaluated definitions overlap to varying degrees, agreement between the good-risk UK-GR profile and MCN 54–67 was consistently high across most trials. Among 712 UKALL2003 HHD-patients, 493 (69%) were positive for both MCN 54–67 and UK-GR, with 5-year EFS of 93.3%; among 843 AIEOP-BFM ALL 2009 HHD-patients, 603 (72%) were positive for both, with 5-year EFS of 91.3% (Supporting Information S2: Table 3). Discordant cases were less favorable: patients meeting UK-GR but with an MCN of 51–53 (7-13% of HHD-cases across trials) had intermediate EFS (e.g., 89.1% in AIEOP-BFM ALL 2009 and 90.7% in UKALL2003) and MCN 54–67 patients who were UK-GR-negative had EFS >78% in all trials (Supporting Information S2: Table 3), but generally below the concordant MCN 54–67/UK-GR group. The principal exception was ALL-IC BFM 2009, in which agreement was substantially lower, possibly reflecting the quality of the cytogenetic analysis and hence the ability to assign the UK-GR profile accurately. Thus, the two definitions identify large, overlapping populations with favorable, but not universally identical, outcomes.

Our findings are consistent with, but not identical to, previous reports. In the St. Jude Children's Research Hospital (SJCRH) analysis of HHD, a DNA-index of 1.16–1.6 remained independently prognostic and was proposed as a clinically practical definition.¹⁸ Our study did not include DNA-index-defined HHD because inclusion of patients identified by different technologies would have introduced diagnostic bias, whereas the SJCRH analysis did not separately assess the MCN 51–53, 54–55, and 56–67 categories.¹⁸ The inferior outcome of MCN 51–53 in our dataset suggests that broad MCN 51–67 groupings can dilute favorable HHD-classifications. Recent genomic studies further show that relapse risk in HHD-ALL is also shaped by cooperating lesions and subtype-specific aneuploidy patterns, underscoring that chromosome-based definitions should ultimately be refined by integrated genomic profiling where available.¹⁹

Several limitations should be emphasized. This retrospective study spans a 20-year period including trials initiated before 2000 as well as more contemporary studies.^6-15 Chemotherapy backbones, risk stratification, MRD methods, cytogenetic review, and supportive care differed between protocols.^6-15 As the study is based exclusively on conventional cytogenetics, it enabled evaluation of the historical definitions as originally proposed and made the analysis relevant to settings where karyotyping remains standard, but it also carries limitations of chromosome morphology, failure rate, and lower resolution compared with SNP-array or next-generation sequencing.¹⁷ Some HHD-cases may have been missed despite uniform eligibility criteria. Because only the eight pre-specified subgroups were collected as aggregate, study-level summary data rather than individual-patient data, we could not test alternative MCN thresholds, perform a formal pooled meta-analysis or treatment-adjusted hazard-ratio comparisons, or compare non-overlapping definitions by formal hypothesis testing. The subgroup comparisons therefore remain descriptive and should be interpreted together with the reported 95% confidence intervals (Supporting Information S2: Table 5). This study was designed as a descriptive, cross-trial reproducibility assessment and is hypothesis-generating rather than a formal test of competing HHD-definitions. Its evidential weight rests on the consistent reproducibility of the same subgroup ordering across eight independently conducted trials, on the basis of which MCN 54–67 and UK-GR are proposed as candidate definitions for prospective evaluation. For the same reason, the pattern of individual chromosomal gains and per-subgroup baseline features of the MCN 51–53 group could not be characterized. This group carries fewer gains by definition and may represent a biologically distinct, lower hyperdiploid subset warranting dedicated study. An i-BFM study aiming to characterize the biology of this lower hyperdiploid subgroup is currently ongoing. Finally, cohorts treated on the ALL-IC BFM 2009 protocol clustered together across subgroups, likely reflecting differences in resources, supportive care, and cytogenetic ascertainment rather than genuine biological heterogeneity. This cohort was retained as a conservative, real-world estimate rather than excluded.

The clinical interpretation must also consider the modern evolving treatment landscape. Blinatumomab has improved outcomes in standard-risk childhood B-ALL and may reduce the relative discriminatory value of classical cytogenetic classifiers.⁵ Recent SJCRH trials XV/XVI reporting 5-year EFS of 94.7% in HHD-patients likewise indicate that outcomes in contemporary protocols may exceed those observed in historical cohorts.^18-20 Therefore, MCN 54–67 and UK-GR should not be used alone to justify therapy reduction. Instead, they should be viewed as practical baseline genetic markers to be integrated with modern molecular diagnostics and MRD, and prospective protocol-specific validation.^{17, 20}

In conclusion, MCN 54–67 and the UK-GR profile consistently defined large, readily identifiable HHD-B-ALL subgroups that ranked among the most favorable across this multinational dataset, although their performance varied across protocols. Conversely, MCN 51–53 appears to represent a smaller HHD-subset that has inferior relative outcome in this context and should not be considered suitable for treatment de-intensification but requires further biological characterization. Our findings support prospective evaluation of MCN 54–67 and UK-GR as components of integrated low-risk stratification, particularly as reference benchmarks for MRD- and immunotherapy-based trials, rather than as standalone classifiers for treatment de-escalation.

Andishe Attarbaschi: Conceptualization; writing—original draft; methodology; writing—review and editing; supervision. Amir Enshaei: Conceptualization; writing—original draft; methodology; visualization; writing—review and editing; data curation. Saskia Sonnenberg: Writing—review and editing; resources; validation. Chloé Arfeuille: Writing—review and editing; resources; data curation; validation. Draga Barbaric: Writing—review and editing; resources. Lena Behrman: Writing—review and editing; resources. Anke Bergmann: Writing—review and editing; methodology; validation; resources; data curation. Nicole Bodmer: Writing—review and editing; resources. Judith M. Boer: Methodology; validation; writing—review and editing; resources; data curation. Myriam Campbell: Writing—review and editing; resources; funding acquisition. Gunnar Cario: Writing—review and editing; resources; conceptualization. Hélène Cavé: Conceptualization; writing—review and editing; methodology; validation; resources. Lydia Champezou: Writing—review and editing; resources. Sarah Elitzur: Writing—review and editing; resources; investigation. Gabriele Escherich: Methodology; validation; writing—review and editing; resources. Hester A. de Groot-Kruseman: Methodology; validation; writing—review and editing; data curation; resources. Monique L. den Boer: Methodology; validation; writing—review and editing; data curation; resources. Csongor Kiss: Funding acquisition; writing—review and editing; formal analysis; resources; data curation. Jelena Lazic: Writing—review and editing; resources; formal analysis. Anja Möricke: Conceptualization; writing—review and editing; methodology; validation; data curation; resources. Hannah Muhle: Writing—review and editing; resources; formal analysis. Karin Nebral: Methodology; validation; writing—review and editing; formal analysis; resources; data curation. Rob Pieters: Funding acquisition; writing—review and editing; formal analysis; resources. Pierre Rohrlich: Writing—review and editing; resources. Martin Schrappe: Funding acquisition; conceptualization; writing—review and editing; resources. Lucie Srámková: Writing—review and editing; resources; investigation. Ajay Vora: Writing—review and editing; resources. Emese Zsigrai: Methodology; validation; formal analysis; resources; writing—review and editing. Arndt Borkhardt: Conceptualization; writing—original draft; writing—review and editing; formal analysis; resources; supervision. Martin Zimmermann: Conceptualization; writing—original draft; writing—review and editing; methodology; validation; formal analysis; supervision; resources. Anthony V. Moorman: Conceptualization; investigation; writing—original draft; writing—review and editing; methodology; formal analysis; supervision.

A.A.—honoraria: Jazz Pharmaceuticals, Amgen, and Novartis; consulting or advisory role: Jazz Pharmaceuticals, Amgen, Novartis, and Takeda Science Foundation; travel, accommodations, expenses: Jazz Pharmaceuticals. S.S.—consulting and advisory role: SERB Pharmaceuticals, Jazz Pharmaceuticals. G.C.—research support: Amgen, Clinigen, Jazz Pharmaceuticals, and Servier; travel, honoraria: Amgen, Clinigen, and Jazz Pharmaceuticals. S.E.—consulting or advisory role: Jazz Pharmaceuticals. A.M.—travel: Jazz Pharmaceuticals; consulting or advisory role: Jazz Pharmaceuticals. M.S.—research support: Shire, Clinigen, Jazz Pharmaceuticals, Servier, and Amgen; honoraria: Servier, Amgen, Clinigen, and Jazz Pharmaceuticals. All other authors declare no conflicts of interest.

Trials were conducted according to the Declaration of Helsinki and with approval of the respective ethics committees.

A.A. and K.N. were supported by St. Anna Children's Cancer Research Institute; S.S., G.C., A.M., M.S., and M.Z. were supported by Deutsche Krebshilfe e.V. (funding number 108588) for the AIEOP-BFM ALL 2009 trial; M.C. was supported by Chilean National Pediatric Cancer Program; R.P. and M.L.d.B. received core funding from KiKa and J.M.B. was supported by KiKa grant 264; L.C.'s fellowship at the EORTC Headquarters was supported by the EORTC Cancer Research Fund (ECRF); C.K. and E.Z. were supported by NKFIH-OTKA K143021 and by Foundation for Children with Leukemia (Leukemias Gyermekekert Alapitvany).

The data that support the findings of this study are available from the corresponding author upon reasonable request.