Comparative survival outcomes between therapy-related and de novo acute lymphoblastic leukemia in adults: a systematic review and meta-analysis

Abstract

Background

Therapy-related acute lymphoblastic leukemia (t-ALL) is an aggressive subtype of ALL that arises after cytotoxic therapy. It is associated with adverse cytogenetics and older age, but survival outcomes compared with de novo ALL remain uncertain. We aimed to systematically evaluate survival differences in adults with t-ALL versus d-ALL.

Methods

We performed a systematic review and meta-analysis according to PRISMA guidelines. PubMed, EMBASE, Scopus, Web of Science, and Cochrane Library were searched to Dec 31, 2024. Eligible studies included adults (≥ 18 years) with t-ALL or de novo ALL reporting survival. Pediatric studies, case reports, reviews, and abstracts were excluded. Two reviewers independently screened and extracted data. Quality was assessed with the MASTER scale. Hazard ratios (HRs) for overall survival were pooled using Doi’s quality effects model. The protocol was registered with PROSPERO (CRD42025625294).

Results

From 3,325 records, 27 retrospective cohort studies (169,237 patients; 1,827 with t-ALL) were included. Nine studies (30,527 patients) were eligible for meta-analysis. The pooled HR for mortality in t-ALL versus d-ALL was 1.07 (95% CI 0.94–1.23), showing no significant survival difference. Median survival ranged from 6 to 32 months in t-ALL and 11–50.6 months in d-ALL. Poor-risk cytogenetics were more frequent in t-ALL (43–100% vs. 31–66%), with higher TP53 mutations (38% vs. 10%) and complex karyotypes. Complete remission was lower in t-ALL (60–88.9% vs. 81.5–93%). Heterogeneity was moderate (I²=50.5%) with evidence of small-study effects.

Conclusion

Despite adverse biology, t-ALL demonstrates survival comparable to d-ALL with modern therapies, particularly allogeneic transplantation.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-025-15366-4.

Introduction

Therapy-related acute lymphoblastic leukemia (t-ALL) is a distinct and challenging subtype of acute lymphoblastic leukemia (ALL) that arises as a complication of previously treated malignancies through either chemo or radiation therapy [1]. Unlike de novo ALL (d-ALL), which occurs spontaneously without prior exposure to cytotoxic treatments, t-ALL exhibits unique genetic and cytogenetic features and is associated with a poorer prognosis [1, 2].

Typically, t-ALL is associated with older age, unique cytogenetic abnormalities such as MLL gene rearrangements, also known as KMT2A, and a history of treatment for solid or hematologic cancers. This condition can arise after a latency period of several years, depending on the type of prior therapy and involved genetic factors [2]. Despite its distinct biology, the clinical outcomes of t-ALL are comparable to d-ALL in patients who undergo allogeneic hematopoietic cell transplantation, emphasizing the need for optimized treatment approaches and early diagnosis [2]. The frequency of t-ALL is not as common, based on earlier studies, it accounted for 1.2–2.5% of adult ALL cases [3, 4], and 4.7% in the largest analysis from the Surveillance Epidemiology and End Results (SEER) program [5].

The terms secondary ALL and t-ALL are often used interchangeably in the literature, yet they are not identical. t-ALL specifically refers to cases that develop following prior exposure to cytotoxic chemotherapy or radiotherapy, whereas secondary ALL describes any ALL occurring after a previous malignancy, regardless of treatment history [2]. This overlap complicates the precise identification of clinical and genetic features unique to t-ALL. While large registry studies on secondary ALL have highlighted poorer survival outcomes, they often do not differentiate between cases linked to prior malignancies without cytotoxic treatment and those directly caused by such therapies [5, 6]. Moreover, these studies are limited by the lack of detailed information on the genetic characteristics of ALL and the specific therapies used for prior cancers due to the constraints of registry data [5, 6]. Additionally, despite new therapeutic agents being introduced, there remains be difficulty experienced by the adult population manifesting as high levels of both treatment-related toxicities and relapses [2, 7].

Thus, this meta-analysis aims to compare the survival outcomes between t-ALL and d-ALL.

Methods

Study design and protocol registration

This study is a systematic review and meta-analysis of studies performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Supplementary Table S1) [8]. The study protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO) with registration ID CRD42025625294.

Search strategy and study selection

PubMed, EMBASE, Scopus, Web of Science, and Cochrane Library were searched for articles published until 31 December 2024 using MeSH terms and keywords related to ALL and therapy-related. No language or date restrictions were applied. A manual reference search was also conducted to ensure no relevant articles were missed. The complete search strategy is in the supplementary materials (Supplementary Tables S2–S6). Articles were imported into EndNote 20 for duplicate removal, then uploaded to the Rayyan systematic review management website (https://www.rayyan.ai/). Within Rayyan, two reviewers (R.M., R.H.) independently screened each study for inclusion based on title and abstract. Disagreements between the two reviewers were resolved through discussion. Abstracts available in languages other than English were translated using Google Translate and then screened for eligibility. Eligible articles in the title and abstract screening were retrieved and underwent full-text screening by the same reviewers. The numbers of articles in each stage were documented in a PRISMA flowchart [9].

Inclusion and exclusion criteria

According to the PICO principle, the inclusion criteria of the articles were as follows:

(1) population: Adult patients (18 years or older) with ALL, (2) exposure: patients diagnosed with t-ALL, (3) control: patients with d-ALL, (4) outcome: (e.g., overall survival (OS), Event-free survival (EFS), and Progression-free survival (PFS)). Cohort studies investigating the survival outcomes in adults above 18 years of age, comparing survival outcomes between t-ALL with d-ALL, were included. Studies were excluded if the study population was less than 18 years of age. Studies were also excluded if they were case series, case reports, letters, opinions, narrative reviews, or letters, opinions, conference abstracts, dissertations, or other non-original articles. Studies without primary data or with incomplete information about study objectives were also excluded.

Outcomes

The primary outcome of this meta-analysis was overall survival, measured as the hazard of death in patients with t-ALL compared to those with d-ALL. Follow-up across the included studies commenced at the time of ALL diagnosis.

Data extraction

For each included study, two reviewers (R.M. and R.H.) independently extracted data using a standardized data extraction sheet created in Microsoft Excel. Extracted data included study characteristics (author, year, country, follow-up period, study design, and sample size), participant characteristics (age, gender, type of prior malignancy, prior therapies), and outcome data relevant to the meta-analysis. Specifically, we extracted hazard ratios (HRs) and their corresponding 95% confidence intervals (CIs) for survival outcomes in both groups. Additional data required for risk of bias and quality assessment were also collected. Discrepancies between reviewers were resolved through discussion and consensus. De-identified study-level data and analysis code are available on request to the corresponding author.

Assessment of the quality of included studies

The quality of the included studies was assessed using the Methodological Standard for Epidemiological Research (MASTER) scale V1.01, which consisted of 36 safeguards under seven methodological standards (Supplementary Table S7). These domains were (1) equal recruitment, (2) equal retention, (3) equal ascertainment, (4) equal implementation, (5) equal prognosis, (6) sufficient analysis, and (7) temporal precedence [10]. Quality assessment was determined by two reviewers (H.K. and Y.E.) independently, and any disagreements were resolved by discussion. The assessment was quantified by enumerating the number of safeguards out of 36 implemented in each study and a relative rank [11]. Based on this count was used in the bias adjustment of the meta-analysis.

Synthesis of findings

We narratively described qualitative data, including author names, year of publication, country, study design, and sample size in tables. For the time-to-event outcome of the hazard ratio of death between t-ALL and d-ALL, the hazard ratio with its 95% confidence intervals was synthesized using the bias-adjusted quality effects model [12]. The latter is one of the models under the common parameter’s assumption and accounts for the heterogeneity between the studies through a quasi-likelihood method. The model also requires a third input (compared to other models whose input is the study effect and its standard error), and this is the bias assessment information in the form of relative ranks to enable bias adjustment. The bias adjustment works by redistributing inverse variance weights towards studies of higher relative rank. This is in distinction to the random effects model, a model under the exchangeable parameters assumption that redistributes weights only in one direction (from larger to smaller studies) and has a larger overall error than the models under the common parameters assumption with decreased coverage of the confidence interval and overconfident results [13–16]. The metan package in Stata was used to perform meta-analysis utilizing study-reported HR and their confidence intervals to compute log-transformed HR’s and their standard errors [17]. To present the data on the HR scale, the synthesized log-HRs were exponentiated for reporting. Heterogeneity was assessed through the I² statistic with I² cut-offs of 25%, 50%, and 75% to classify heterogeneity as low, moderate, or high, respectively [18].

Publication bias was assessed through the Doi plot and LFK index [19]. Sensitivity analyses were conducted to assess the robustness of the results, involving excluding each study and assessing its impact on the overall estimate. All analyses were conducted using Stata 18.5.

Results

Search results

Figure 1 provides an overview of the search process. A total of 3325 study reports were identified from the database and manual searches. Out of these, 2207 duplicates were removed. The remaining 1118 records were imported into Rayyan for the title and abstract screening. Out of these, 137 studies remained for full-text screening. One hundred and ten studies were excluded during this stage for the reasons outlined in the PRISMA flowchart (Fig. 1) and in (Supplementary Table S8). A total of 27 studies met the eligibility criteria and were included in the systematic review. Of these, 9 studies were included in the meta-analysis. The remaining 18 studies were excluded from the meta-analysis due to reasons such as insufficient quantitative data, heterogeneity in outcome measures, or methodological differences that precluded meaningful statistical pooling.

Fig. 1.

PRISMA flow diagram

Characteristics of included studies

A total of 27 studies comprising 169,237 participants were included in this systematic review, with 1827 cases of t-ALL. A total of 9 studies enrolling 30,527 participants were included in the meta-analysis. All studies (27) utilized a retrospective cohort design. These studies were conducted across 25 countries or centers, reflecting broad geographical diversity. The underlying malignancies leading to t-ALL were predominantly breast cancer, which was the most frequently reported prior cancer. In terms of immunophenotypic characterization, the majority of t-ALL cases exhibited a B-cell phenotype, with an average prevalence of 89.0%. Most studies reported clinical outcomes, including complete remission (CR) rates, OS at 12 and 24 months, with variable follow-up durations. Hematopoietic cell transplantation (HCT) was frequently utilized, although the proportion of patients receiving HCT varied significantly across studies. The detailed characteristics of the 27 studies and their main findings are summarized in Table 1.

Table 1.

Characteristics of all included studies

Study Author	Year	Center/Country	Cohort years	Study design	Total Number of participants	No. of participants with t-ALL	Male (%)	Median age (year)	Median Gap duration (months)	B cell phenotype (%)	CR (%)	No. of patients received HCT	Median OS (months)	Median EFS (months)	Type	Induction Regimens	Transplant Eligibility Criteria
Pagano [3]	1999	GIMEMA Archive (Multi-center), Italy	1992–1996	Retrospective cohort	3934	21	57.14	58	27	95.2	57		5			ACs ± cyclophosphamide, VCR + PD
Bloomfield [20]	2002	United States, Europe, Japan International (19 centers)	1993–2002	Retrospective cohort	162	13	38.46	45	26			5	8			TIs, AAs, AMs, L-asparaginase, VCR, PD	Not Specified
Casorelli [21]	2003	Division of Hematology of the Catholic University of Rome, Italy	1992–2001	Retrospective cohort	25	4	25	58	ranged from12-84							AAs, ACs, AMs, TIs, Cisplatin, Paclitaxel, Tamoxifen, VCR, Bleomycin	Not Specified
Ishizawa [4]	2003	City of Hope National Medical Center, USA	1989–2001	Retrospective cohort	152	6	33.33	59.5	ranged from 13–96	100	33		17.45	37.5	DFS	TIs, AAs, AMs, Paclitaxel, fludarabine, dexamethasone	Not Specified
Shim [22]	2010	Asan Medical Center, South Korea	1989–2009	Retrospective cohort	9228	3	0	55	25	67	25			27	DFS	TIs, AAs, Cisplatin, Paclitaxel, 5-Fluorouracil
Abdulwahab [23]	2012	Princess Margaret Hospital, Canada	1989–2011	Retrospective cohort	335	23	34.78	51	48	91	81	5	13.6	27	DFS	DFCI, Modified DFCI, HyperCVAD, Imatinib	Not Specified
Tang [24]	2012	University of Texas MD Anderson Cancer Center	2004–2010	Retrospective cohort	457	30	50	64	36	100	60	3	12			HyperCVAD, TKIs	Not Specified
Ganzel [25]	2015	Memorial Sloan Kettering, USA	1994–2013	Retrospective cohort	807	23	56.52	55	84	87	86	6	13.6			ALL-2, L-20, HyperCVAD, NY-II, MRC UKALL XII/ECOG 2993, VCR + steroids, TKIs, Others	Not Specified
Giri [6]	2015	SEER Database, USA	1992–2011	Retrospective cohort	4203	79	86.08	62	60	90			8			Not Specified	Not Specified
Churpek [26]	2016	University of Chicago, USA	1972–2012	Retrospective cohort	88	7	0	52	58				13			Not Specified	Not Specified
Xu [27]	2016	China	2002–2012	Retrospective cohort	35	4	25	50	24		75		14			VICLP, VIP, VDLP, VILP	Not Specified
Kelleher [28]	2016	Spain	2006–2012	Retrospective cohort	448	16	31.25	55	84	88	94	2				Risk-adapted PETHEMA protocols	Not Specified
Fabiani [29]	2017	Italy	2003–2014	Retrospective cohort	14	1	0	40	18	100	100	1	54			AIDA 2000 protocol, Clofarabine + Cyclophosphamide, LAL0904 protocol	Not Specified
Swaika [5]	2018	SEER Database, USA	1973–2012	Retrospective cohort	11,728	772	50.9	59	60				6			Not Specified
Ferraro [30]	2018	Washington University in St. Louis, USA	2006–2017	Retrospective cohort	211	30	30	57	80	79	79	15	11.8			HyperCVAD	Not Specified
Aldoss [2]	2018	USA (City of Hope)	2000–2017	Retrospective cohort	1022	93	43	55	82	91	85	49	22			HyperCVAD ± TKIs	Not Specified
Saygin [1]	2019	Multi-institutional (5 centers)	2000–2017	Retrospective cohort	1785	116	48.28	61	60	92	87	42	27	18	RFS	BFM-based protocols, HyperCVAD	Not Specified
Nampoothiri [31]	2020	Princess Margaret Cancer Center, Canada	1999–2019	Retrospective cohort	18	18	50	44	47	89	89	18	21.2			Modified Dana-Farber Cancer Institute protocol, HyperCVAD	Not Specified
Abdel Rahman [7]	2021	Mayo Clinic, USA	2008–2019	Retrospective cohort	331	69	33.33	63	60	100	79.4	34	30			HyperCVAD, Others	Not Specified
Kook [32]	2022	Yonsei University, Korea	2012–2021	Retrospective cohort	171	9	55.56	56	77	100	89	4	12.7	10.4	PFS	HyperCVAD ± TKIs, Pediatric GRAALL regimen, VPD regimen, Imatinib	Not Specified
Jabagi [33]	2020	Nationwide, France	2007–2015	Retrospective cohort	122,373	36	0	55	60							Not Specified
Vasudevan Nampoothiri [34]	2022	Princess Margaret Cancer Center, Canada	2001–2021	Retrospective Cohort	20	6	66.67	62.5	75	100	100	6	24			Modified Dana-Farber protocol	Not Specified
Geyer [35]	2023	MSKCC and OSU, USA	2014–2022	Retrospective cohort	32	32	43.75	65	53.8	100	88.9	13	35.8	35.4	EFS	Hyper-CVAD ± rituximab, POMP, Blinatumomab & inotuzumab ozogamicin, CALGB 10,102, targeted therapy	Not Specified
Elghawy [36]	2024	University of Pennsylvania, USA	2010–2023	Retrospective cohort	1770	9	33.33	62.7	60	100	56	0	19.5			Mini-CVD + inotuzumab, HyperCVAD, VCR, Blinatumomab	Not Specified
Sadowska-Klasa [37]	2024	EBMT Registry (Multi-center)	2001–2021	Retrospective cohort	9720	351	32.5	51.8	60	80	100	80	32			Not Specified	≥ 18 yrs old with ALL reported as a secondary malignancy, who underwent the first HCT in CR1 (intermediate disease risk index)
Hoferet [38]	2024	University Hospital Zurich	2009–2019	Retrospective cohort	131	19	52.63	60	78	79						Not specified	Not Specified
Pourhassan [39]	2024	City of Hope, USA	2000–2023	Retrospective cohort	37	37	0	57	79	92	92	21	19.4			HyperCVAD, BFM-based therapy, Blinatumomab, Inotuzumab ozogamicin, TKIs	Not Specified

GIMEMA Gruppo Italiano Malattie EMatologiche dell’Adulto, SEER Surveillance, Epidemiology, and End Results database, MSKCC Memorial Sloan Kettering Cancer Center, OSU Ohio State University, EMBT European Society for Blood and Marrow Transplantation registry, CR Complete Remission Rate, OS Overall Survival, EFS Event-Free Survival, PFS Progression-Free Survival, DFS Disease-Free Survival, RFS Relapse-Free Survival, AA Alkylating agents, ACs Anthracyclines, AMs Antimetabolites, TIs Topoisomerase II inhibitors, VCR Vincristine, PD Prednisone, DFCI Dana-Farber Cancer Institute pediatric-based protocol, TKIs Tyrosine Kinase Inhibitors

Assessment of the quality of included studies

The included studies were generally of moderate quality, with the number of safeguards against systematic error ranging from 23 to 26, averaging 24.7 out of 36 safeguards. Most studies demonstrated strong safeguards in key domains assessed by the MASTER scale, including equal recruitment, equal retention, sufficient analysis, and good temporal precedence. However, common deficiencies were observed in the domains of equal ascertainment (particularly regarding blinding of analysts), equal implementation (notably in ensuring comparable co-interventions), and equal prognosis (specifically in random allocation and concealment of allocation procedures). The individual assessments for all the studies are shown in the supplementary materials (Supplementary Table S7).

Descriptive analysis of the included studies in the meta-analysis

Table 2 presents a summary of key findings from studies on both d-ALL and t-ALL, including phenotype distribution, prior treatments, cytogenetic abnormalities, hazard ratios, and survival outcomes. Across nine studies, up to 28,045 participants were included in the meta-analysis. B-cell ALL was significantly more common than T-cell ALL, with proportions ranging from 79% to 100% and 0% to 21%, respectively. Prior chemotherapy exposure was reported in 38% to 85% of patients, while radiotherapy was observed in 11% to 56%, and combined chemoradiotherapy in up to 40%.

Table 2.

Ranges of outcomes reported across studies

Variable	t-ALL (Range/%)	de novo ALL (Range/%)	Participants (n)	No. of Studies
Clinical features
Prior chemotherapy (%)	38–85	—	13,697	7
Prior radiotherapy (%)	11–56	—	13,697	7
Prior chemoradiotherapy (%)	0–40	—	3,977	6
HCT recipients (%)	10–100	50–100	698/10,660	7/4
Complete remission (%)	60–88.9	81.5–93	347/2,047	6
Cytogenetics
TP53 mutation (%)	38	10	9/162	1
Complex karyotype (%)	5.8–11.1	3.1–5.7	78/424	2
Other translocations (%)	3.3–23	0.6–25	338/1885	5
Standard-risk cytogenetic (%)	13–43	16.4–51	698/11,416	7
Poor-risk cytogenetic (%)	43–100	31–66	698/11,416	7
Phenotype
B-cell phenotype (%)	80–100	—	747	7
T-cell phenotype (%)	0–20	—	747	7
Outcomes
HR (tr-ALL vs. d-ALL)	1.04–1.72		28,045	9
HR (B-cell vs. T-cell phenotype)	1.3–1.59		5441	3
HR for recurrence	0.97–1.24	—	10,051	2
Median OS (months)	6–32	11–50.6	28,045	9
HR for mortality after HCT (HCT vs. non-HCT)	0.41–1.29	—	2012	4
HR for EFS	1.09–1.39	—	9931	2

Regarding cytogenetics as presented in Fig. 2, TP53 mutations were notably more frequent in the t-ALL group (38%) compared to the de novo group (10%), based on data from a single study. Complex karyotypes, defined as the presence of three or more chromosomal abnormalities in the absence of a single defining translocation, were also more prevalent in t-ALL (5·8–11·1%) compared to d-ALL (3.1–5.7%). Other cytogenetic translocations, including but not limited to MLL rearrangements and t (9;22) (BCR-ABL), were reported in 3.3% to 23% of t-ALL cases and 0.6% to 25% of d-ALL cases across five studies. Cytogenetic risk was categorized based on established prognostic markers: poor-risk cytogenetics included abnormalities such as complex karyotypes, TP53 mutations, hypodiploidy, and MLL rearrangements, while standard-risk cytogenetics encompassed normal karyotypes or favorable rearrangements like hyperdiploidy. Poor-risk cytogenetics were notably more frequent in t-ALL (43–100%) than in d-ALL (31–66%), while standard-risk profiles were less frequent in t-ALL (13–43%) compared to de novo cases (16.4–51%).

Fig. 2.

Distribution of Major Cytogenetic Abnormalities Between Therapy-Related and De Novo Acute Lymphoblastic Leukemia (ALL)

The median follow-up for mortality ranged from 6 to 32 months in t-ALL and 11 to 50.6 months in d-ALL. Hazard ratios for various treatment comparisons were reported. Saygin et al. reported a significantly increased mortality risk associated with radiotherapy compared to chemotherapy, with an HR of 8.25 (95% CI: 1.06–64.07) [1]. Two studies reported HRs for recurrence, ranging from 0.97 to 1.24 [32, 38]. Across studies, HCT-related outcomes showed variable mortality benefit when comparing patients who underwent transplantation versus those managed without HCT within the t-ALL cohort (HR 0.41–1.29). PFS was reported in one study, with a median of 10.4 months in the t-ALL group compared to 13.4 months in the d-ALL group [32].

Meta-analytic estimates

Figure 3 presents a forest plot illustrating the individual hazard ratio estimates and their respective weights derived using Doi’s Quality Effects model. The pooled estimate of the hazard ratio was 1.07 (95% CI: 0.94–12.3), indicating no statistical divergence. Among the individual studies, statistically divergent HRs were reported from two studies (HR = 1.54, 95% CI: 1.16–2.04) and (HR = 1.12, 95% CI: 1.03–1.22) as reported in Giri et al. and Swaika et al., respectively [5, 6]. In contrast, the remaining studies yielded hazard ratios with narrower confidence intervals that included the null value, suggesting no significant difference between the two groups. The heterogeneity was moderate with an I² of 50.5% demonstrating a moderate level of inconsistency across included studies. Publication bias was assessed using the Doi plot and the LFK index and showed major asymmetry (Supplementary Fig. 1).

Fig. 3.

Forest plot of the hazard of mortality in the two ALL variants

Sensitivity analysis

The sensitivity analysis was performed and suggested that the pooled effect size is robust as omitting any single study does not substantially change the overall estimate. However, one study showed a noticeable shift, indicating potential influence on the overall estimate (Supplementary Fig. 2) [32].

Prevalence of Therapy-Related ALL across studies

Figure 4 displays a forest plot summarizing the prevalence of t-ALL across nine included studies. The estimated proportions of t-ALL varied across studies, ranging from as low as 0.02 (95% CI: 0.02–0.02) in Giri et al. [6] to as high as 0.14 (95% CI: 0.10–0.20) in Ferrao et al. [30]. Most individual study estimates clustered between 0.04 and 0.07.

Fig. 4.

Forest plot of the prevalence of t-ALL across studies

The pooled prevalence of t-ALL, calculated using Doi’s Quality Effects model, was 0.04 (95% CI: 0.01–0.07), indicating an overall low prevalence. Notably, the between-study heterogeneity was substantial, with an I² of 97.8% (p < 0.001), reflecting considerable variability in the reported proportions.

Remission after initial therapy

Figure 5 presents a pooled analysis of odds ratios comparing the likelihood of achieving complete remission after initial therapy between t-ALL and d-ALL. The meta-analysis yielded a pooled odds ratio of 0.54 (95% CI: 0.19–1.54), suggesting a trend toward lower remission rates in the t-ALL group; however, the association was not statistically significant. Between-study heterogeneity was substantial (I² = 81.1%, p < 0.001), indicating notable variability in effect estimates across studies. Among the individual studies, Tang et al. [34] reported a significantly lower odds of remission in t-ALL (OR = 0.11, 95% CI: 0.05–0.26), while Ferraro et al. [30] also demonstrated reduced odds, though not statistically significant (OR = 0.44, 95% CI: 0.16–1.22). Other studies, including Saygin [1] and Kook [32], reported wide confidence intervals encompassing the null, indicating no statistically significant difference.

Fig. 5.

Forest plot of the remission after initial therapy between t-ALL and de novo

Discussion

This systematic review and meta-analysis investigated survival outcomes in adult patients with t-ALL compared to those with d-ALL. The pooled hazard ratio of 1.07 (95% CI: 0.94 − 1.23) indicates no statistically significant difference in overall survival; however, this finding warrants cautious interpretation. The apparent similarity in outcomes may be partly attributable to improvements in transplant availability, supportive care, and treatment intensity in recent years. Moreover, most included studies were retrospective and likely favored younger, fitter t-ALL patients selected for curative therapy, potentially minimizing true prognostic disparities. Future prospective research using uniform treatment algorithms and cytogenetic stratification is essential to validate whether therapy-related ALL truly parallels de novo ALL in survival outcomes. While the pooled hazard ratio did not demonstrate a statistically significant difference in survival, the upper confidence limit suggests a possible 23% higher mortality risk in therapy-related ALL compared with de novo ALL. Thus, the findings indicate broadly comparable but not identical outcomes, and a modest survival disadvantage for therapy-related cases cannot be ruled out.

From the 27 included studies, nine of which were meta-analyzed, distinct patterns emerged. Notably, B-cell immunophenotype predominated in t-ALL cases, ranging from 80% to 100% across seven studies. This aligns with prior findings that therapy-related leukemias frequently exhibit pro-B (CD10-negative) features, particularly in cases harboring KMT2A (MLL) rearrangements, a hallmark of therapy-induced leukemogenesis [1, 40]. Such rearrangements are typically associated with aggressive disease, chemotherapy resistance, and poor remission rates, especially when coupled with an early immature B-cell phenotype [4, 40].

One of the most compelling differences between t-ALL and d-ALL is the cytogenetic risk distribution. According to the descriptive table, poor-risk cytogenetics were more frequent in t-ALL (43–100%) compared to d-ALL (31–66%). Moreover, standard-risk cytogenetics were less frequent in t-ALL (13–43%) versus de novo (16–51%), suggesting a skewed genetic risk profile toward more aggressive disease in t-ALL. Notably, t-ALL patients also exhibited a higher prevalence of TP53 mutations (38% vs. 10%) and complex karyotypes (5.8–11.1% vs. 3.1–5.7%), both of which are independently associated with resistance to chemotherapy, rapid relapse, and poor survival outcomes [2, 41, 42]. These cytogenetic features likely reflect the underlying clonal hematopoiesis and DNA damage accumulated from prior exposure to alkylating agents or topoisomerase II inhibitors.

Despite these high-risk molecular characteristics, CR rates in t-ALL ranged from 60 to 88.9%, which, although slightly lower than in d-ALL (81.5–93%), suggests that a significant proportion of t-ALL patients can still achieve remission with intensive therapy. Importantly, across the included studies, HCT was more frequently utilized in t-ALL (10–100%, 7 studies) compared with d-ALL (50–100%, 4 studies). This disparity suggests that clinicians may pursue transplantation more aggressively in t-ALL, either due to upfront risk stratification or early relapse patterns. Prior work, including Aldoss et al., supports this approach, demonstrating that HCT can overcome the adverse prognostic impact of therapy-related disease in selected patients [2].

Furthermore, the duration of OS also showed a marked difference: the median OS in t-ALL ranged from 6 to 32 months, whereas in d-ALL, it ranged from 11 to 50.6 months. While not statistically significant in the pooled model, this trend remains clinically meaningful, particularly in lower-income settings or populations with limited access to transplant and advanced therapies.

The apparent paradox, comparable pooled survival despite lower CR rates, shorter median OS, and a higher burden of poor-risk cytogenetics, underscores the influence of treatment intensity and selection bias across the included studies. The majority of t-ALL patients represented in these cohorts were younger, had fewer comorbidities, and were eligible for intensive induction regimens or allogeneic transplantation. This selected subgroup likely achieved outcomes that offset the adverse biological risk associated with therapy-related disease. Conversely, patients with poor functional status, extensive prior therapy, or multiorgan impairment were underrepresented and may have contributed disproportionately to early mortality in real-world practice. Therefore, the observed “no significant difference” in pooled survival should not be interpreted as biological equivalence between t-ALL and de novo ALL, but rather as a reflection of modern treatment success in fit, transplant-eligible patients.

Another observation from the descriptive data concerns the history of prior therapy: 38–85% of t-ALL patients had received chemotherapy, and 11–56% had received radiotherapy. This reinforces the mechanistic link between t-ALL and genotoxic exposure, further evidenced by the predominance of MLL rearrangements and TP53 mutations in these patients. These mutations often reflect clonal evolution from prior malignancy treatment, consistent with the latency periods (often 1–5 years) documented in the literature [1, 41]. Lastly, the variation in HRs for survival across studies (ranging from 1.04 to 1.72) suggests that not all t-ALL patients carry a uniformly poor prognosis. Some cohorts, particularly those with access to transplant, modern induction regimens, or MRD-guided therapy, achieve outcomes comparable to their de novo counterparts. However, it was demonstrated that in large, registry-based cohorts, t-ALL is independently associated with worse survival, reinforcing the need for early aggressive management and tailored therapeutic strategies as reported in Giri et al. and Swaika et al. [5, 6].

This meta-analysis has several notable strengths. First, it represents the largest synthesis to date of survival outcomes in t-ALL, incorporating 27 retrospective cohort studies with a combined population of 169,237 patients, including 1827 with t-ALL. The inclusion of nine studies with 30,527 patients in the quantitative meta-analysis provided robust statistical power. Importantly, the study was conducted in accordance with PRISMA guidelines and was prospectively registered on PROSPERO, ensuring methodological transparency and adherence to international standards. The use of Doi’s Quality Effects model, which incorporates study quality into effect estimates, further strengthens the reliability of the pooled findings [12]. Another important strength lies in the detailed cytogenetic and molecular insights provided. Clinically, the findings are highly relevant: despite the enrichment of adverse biological features in t-ALL, survival outcomes were not significantly different from d-ALL when modern treatment approaches, including HCT, were employed.

However, several limitations must be acknowledged. All included studies were retrospective, introducing the possibility of selection bias and incomplete reporting. Although cytogenetic data were available in many studies, detailed molecular profiling was lacking in several, with TP53 mutation analysis, for example, reported in only one study. Hence, future prospective studies with standardized molecular profiling and harmonized reporting are essential to better delineate subtype-specific outcomes and reduce unexplained heterogeneity.

Another limitation is the relatively short to moderate follow-up period across studies, ranging from 6 to 32 months, which may underestimate late relapses and long-term treatment failures. Additionally, high heterogeneity of treatment modalities was observed among included studies. Patients received different regimens, including chemotherapy, hematopoietic cell transplantation, and, in some cases, targeted therapies. However, most studies did not report survival outcomes stratified by treatment type, precluding a formal subgroup or meta-regression analysis. This variability may have influenced pooled estimates, highlighting the need for future research with standardized, treatment-specific reporting to enhance clinical relevance.

In summary, while the meta-analysis did not reveal a statistically significant survival disadvantage in t-ALL overall, the descriptive and genomic data establish t-ALL as a biologically and clinically distinct high-risk entity. Key features include older age, prior chemotoxic therapy and adverse cytogenetics such as TP53 mutations, complex karyotypes, KMT2A rearrangements. These insights support the need for early molecular profiling, MRD monitoring, and consideration for novel therapy and upfront allogeneic transplantation or enrollment in clinical trials.

Conclusion

While t-ALL displays distinct adverse cytogenetic and molecular features, overall survival appears comparable to d-ALL, particularly when patients receive allogeneic hematopoietic stem cell transplantation. These findings support aggressive treatment approaches in fit patients and call for improved molecular stratification in future studies. Despite its adverse cytogenetic and molecular characteristics, t-ALL demonstrated overall survival comparable to d-ALL, underscoring the value of modern therapies and allogeneic HSCT. These findings highlight the importance of early molecular risk stratification and individualized therapy in managing ALL patients.

Authors’ contributions

All authors contributed significantly to the work reported; gave final approval of the current version; and agreed to be accountable for all aspects of the work.RM: Conceptualization, Methodology, Data Curation, Writing-Original Draft, Writing, Review & Editing, Project administration. MNK: Writing-Original Draft, Methodology, Software, Formal analysis. AS: Writing-Original Draft, Methodology, Software, Formal analysis. RH: Methodology, Data Curation, Writing-Original Draft. HC: Conceptualization, Writing - Review & Editing. YE: Methodology, Writing-Original Draft. HK: Methodology, Writing - Original Draft. AAM, DS, DM: Writing-Review & Editing. SAD: Supervision and Formal analysis. SFM: Conceptualization, Methodology, Writing - Review & Editing, Supervision, Project administration.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.