Survival of patients with acute lymphoblastic leukemia (ALL) younger than 60 years has improved during the past two decades . Data emerging from the surveillance, epidemiology and end results (SEER) database suggest that patients' age serves as a significant prognostic factor that affects clinical outcomes such as overall survival . Adolescents and young adults (AYA) are of special interest in that regard. Survival of patients with ALL in this age group, usually defined as 16–39 years , is inferior to that of pediatric patients and might decrease by ∼ 50% between childhood and early adulthood . Possible explanations for this inferiority include variations in the cytogenetic, molecular, and phenotypic profile of ALL among the different age groups, differences in treating disciplines, psychosocial factors unique to this age group, under-representation of this population in available trials, and diversity in treatment regimens [2–7]. Indeed, pediatric protocols and the current pediatric-inspired regimens, when compared with conventional adult protocols, include more courses of re-induction and intensification, higher cumulative doses of cytostatic drugs and a generally longer maintenance phase . So far, most of the comparative data dealing with treatment strategies of AYA with ALL are nonrandomized and limited by this methodology.
To better clarify and summarize the currently available data we conducted a systematic review and meta-analysis of all comparative trials on AYA patients given chemotherapy with either pediatric-inspired regimens or conventional adult protocols.
Materials and Methods
Data sources and search
We followed the published recommendations for meta-analysis of observational studies  to search the literature, perform the analysis, and report the results. A comprehensive search strategy was employed to identify both published and unpublished studies, with no restriction on language, type of publication, or study years. We searched MEDLINE from 1971 through May 2011 and the proceedings of the relevant conferences (American Society of Hematology, American Society of Clinical Oncology, and European Hematology Association) from 2004 to 2010. We used the following search terms: (adolescent or young adults) and (leukemia or leukaemia [MeSH]) and crossed with a highly sensitive search for trials: prospective or longitudinal or cohort or randomized controlled trial [pt] or controlled clinical trial [pt] or randomized controlled trials [mh] or random allocation [mh] NOT (animals [mh] NOT human [mh]). The references of all identified articles were inspected for more studies.
To be eligible for inclusion in this meta-analysis, publications had to include original data from randomized controlled trials, cohort studies, case-control, or nested case-control studies, which a priori (before starting induction chemotherapy) allocated patients to either pediatric-inspired or a conventional adult regimen. Studies comparing only dose intensifications without allocation to “pediatric-inspired” protocol were excluded. In studies with multiple publications of the same cohort, data from the most recent publication were included.
Two reviewers (RR and OW) inspected the title and when available, the abstract of each reference identified in the search, and applied the inclusion criteria. Where relevant articles were identified, the full article was obtained and inspected independently by the above two reviewers and inclusion criteria applied. In cases of disagreement between the two reviewers, the full article was inspected independently by a third reviewer (LV).
The primary outcome measure was all cause mortality at 3 years and at the longest available follow-up. Secondary outcomes included post-induction complete remission (CR) rate, event free survival (EFS), relapse rate, and nonrelapse mortality (NRM).
For quality appraisal, we implemented the Newcastle-Ottawa scale, a standardized quality assessment scale of cohort and case-control studies . This scale assesses three major parameters: selection of cohort and controls (representativeness of the cohort, selection of nonexposed cohort, and ascertainment of exposure–three points), comparability of the selected cases and controls (one point), and adequacy assessment of outcomes (outcome assessment, duration of follow up, and adequate follow up–three points). Adequacy score was summarized for each study (maximal score seven). Nevertheless, comparability of studies was defined as adequately satisfied when cohort and control were controlled for either age or disease risk since these parameters are considered to have a major role for ALL prognosis and treatment allocation . Adequate follow-up was defined as at least 3 years from treatment allocation. We chose this time point to allow the contradictory efficacy effects and treatment associated toxicities to sum up as the net effect on all cause mortality. Scores ranged from 0 to 7, with higher scores representing better methodology. Sensitivity analysis was applied for studies with a score higher than three and for those with well controlled groups.
For all outcomes, dichotomous data were analyzed by calculating the relative risk (RR) for each trial with 95% confidence intervals (CIs) (Review Manager version 5.1). For all cause mortality, relapse, and NRM, RR < 1 favors the interventional arm (pediatric-inspired regimens). For EFS and post-induction CR rate, RR > 1 favors the interventional arm. For all outcomes, we analyzed the data according to intention-to-treat method in which we included all known events in both nominator and denominator. We assessed heterogeneity of trial results by calculating a chi-square test of heterogeneity and the I2 measure of inconsistency. We used the Mantel–Haenszel fixed-effect model for pooling trial results unless a statistically significant heterogeneity was found (P = 0.10 or I2 = 50%), in which case we used the random-effects model . Heterogeneity was investigated through subgroup and sensitivity analyses as defined above. We analyzed studies that included young patients up to 20 years and studies that included also older patients separately. To look for small studies' effect (which might indicate publication bias) we examined a funnel plot of the primary outcome.
Description of studies and regimens
The literature search identified 2002 publications. Seventeen potentially relevant full text articles were retrieved for further evaluation [11–27]. Of these, nine were excluded [19–27]. Three additional studies were identified by searching conference proceedings [28–30] (Fig. 1).
Eleven studies published between 2003 and 2009 reporting on 2,489 patients, met inclusion criteria for systematic reviews [11–18, 28–30]. None of the studies was a randomized controlled trial. Data regarding demographics and interventions are summarized in Table I.
Table I. Characteristics of Included Studies
Age (years); median (range)
% Patients receiving HCT
Duration of protocol (months)
Duration of follow-up (months)
Cumulative medications' doses
IT MTX (mg or N doses)
Ara-C, cytarabine; ASN, asparaginase; CA, conventional adult; CTX, cyclophosphamide; DNR, daunorubicin; Etop, etoposide; HCT, hematopoietic cell transplantation; IT MTX, intrathecal methotrexate; KU, 1000 units; NR, not reported; PI, pediatric inspired; Pred, prednisone; VCR, vincristine.
Regimens of both pediatric-inspired and conventional adult protocols were different between the studies. Eight nonrandomized studies compared patients given pediatric-inspired regimens with those given conventional adult regimens at the same time period [11–13, 15, 16, 18, 28, 30]. Two studies included patients given pediatric-inspired regimens that were compared with historical controls in which conventional adult protocols were given [17, 29]. In one study, the relevant study period was not reported . Only three studies reported the outcome of patients older than 20 years [17, 18, 29]. Two of them reported outcome of patients up to the age of 55 years [17, 29]. The third study  included patients up to the age of 25 years. This study was not included in the subgroup analyses of patients older than 20 years since the median age of patients was only 17 years.
In three studies [17, 18, 29], age was well compared in the two investigated cohorts. In three studies [11, 12, 15], disease risk according to cytogenetic analysis was well compared and in seven studies [11, 12, 14–18] baseline leukocyte counts were well compared between the two investigated cohorts.
Quality appraisal according to the Newcastle-Ottawa quality assessment scale of cohort studies is shown in Table II. Four studies were graded four or above [11, 13, 15, 17]. In five studies, the two investigated cohorts were well-compared (in either patients' age or disease risk) [11, 15, 17, 18, 29]. In one study , there was a significant difference in the median duration of follow up between patients given pediatric-inspired regimens and patients given conventional adult chemotherapy (70 vs. 44 months).
Table II. Quality Assessment According to the Newcastle-Ottawa Quality Assessment Scale of Cohort Studies
Cumulative dosages of the different drugs in the pediatric and adults' regimens are summarized in Table I. Figure 2 depicts for each medication, the ratio between the doses in the pediatric and adults' regimens, according to the various studies. In most studies, pediatric-inspired protocols compared with conventional adult regimens, used higher median doses of corticosteroids (ratio of 2.2; range 0.5–3.8), vincristine (ratio of 2.5; range 0.9–16) asparaginase (ratio of 2.3; range 1–420), and intrathecal methotrexate (ratio of 1.8; range 0.2–2.8). Conversely, pediatric-inspired regimens had lower median doses of daunorubicin (ratio of 0.8; range 0.6–1.3), cytarabine (ratio of 0.9; range 0.1–1.0), and etoposide (ratio of 0.1; range 0–0.3). Adherence to protocol was reported in two studies [11, 15]. Stock et al. analyzed adherence to protocol by determining the percentage of patients in CR starting maintenance within 30 days , while Boissel et al. looked at the median interval between the data of CR date and the first day of the post-remission therapy . While the median interval between the achievement of CR and the first day of post-remission therapy was shorter in the pediatric-inspired group compared with adult regimens (2 vs. 7 days, 177 patients) , there was a lower percentage of patients starting maintenance treatment within 30 days in the pediatric-inspired group (43% vs. 76% of patients achieving CR) . This suggests a delay in treatment attributed to increased toxicity.
Primary outcome: all cause mortality
There was a statistically significant reduction in all cause mortality in patients given pediatric-inspired regimens compared with patients given the conventional adult protocols at 3-years and at the end of study period (RR 0.58, 95% CI 0.51–0.67, 8 trials, 1,956 patients, and RR 0.59, 95% CI 0.52–0.66, 10 trials, 2,246 patients, respectively), Fig. 3a. The absolute risk reduction for all cause mortality at 3 years was 0.20 and the number needed to treat to prevent one death with pediatric-inspired regimens was 5 (95% CI 4–7).
Sensitivity analysis including studies of higher methodological quality (i.e., only studies that were graded four or above according to the Newcastle-Ottawa scale) showed similar results (RR 0.59, 95% CI 0.52–0.68, 3 trials, 1,424 patients) [11, 15, 17]. Sensitivity analysis including only studies in which the two cohorts were well compared in either age or disease risk also showed similar results (RR 0.58; 95% CI 0.50–0.67, 5 trials, 1,512 patients) [11, 15, 17, 18, 29]. One study with an older population  was an outlier and contributed to heterogeneity. When excluded, no heterogeneity was shown. After censoring this study, the absolute risk reduction was 0.28 and the number needed to treat to prevent one death with pediatric-inspired regimens was 5 (95% CI 4–6).
Two studies recruited patients older than 20 years [17, 29], However, they reported outcomes for the whole group rather than for the population above 20 years separately. Thus, we could not analyze this subpopulation separately. In both studies, there was a statistically significant reduction in all cause mortality in patients given pediatric-inspired regimens compared to those given the conventional adult protocols.
Post-induction CR rate.
There was a statistically significant increase in CR rate in patients given pediatric-inspired regimens compared to patients given conventional adult protocols (RR 1.05, 95% CI 1.01–1.10, I2 = 55%, random effects model, 7 trials, 1,947 patients), Fig. 3b. One study showed similar CR rates for both groups (RR 0.99; 95% CI 0.94–1.04) and was an outlier in this analysis . In this study, conventional adult protocols contained relatively higher cumulative doses of corticosteroids and vincristine, potentially contributing to a better CR rate.
Event free survival (EFS).
There was a statistically significant increase in EFS in patients given pediatric-inspired regimens when compared to patients given conventional adult regimens at 3 years (RR 1.66, 95% CI 1.39–1.99, I2=61%, random effects model, 9 trials, 1,739 patients), Fig. 3c. Heterogeneity was mainly contributed by one study . In this study, patients given pediatric-inspired protocols were younger when compared with patients given conventional adult regimens. Sensitivity analysis including studies in which the two cohorts were well compared in either age or disease risk showed similar results (RR 1.76; 95% CI 1.55–2.00, 4 trials, 1,191 patients) [15, 17, 18, 29].
There was a statistically significant reduction in relapse rate in the pediatric-inspired group compared with the standard-adults group (RR 0.51, 95% CI 0.39–0.66, I2 = 54%, random effects model, 8 trials, 1,952 patients), Fig. 3d. Heterogeneity was mainly contributed by one study .
Nonrelapse mortality (NRM).
No difference was shown in the rate of NRM between the two groups (RR 0.53, 95% CI 0.19–1.48, I2 = 56%, random effects model, 4 trials, 436 patients), Fig. 3e. After exclusion of one trial  in which the two groups were statistically significantly different in age (median 15 years in pediatric-inspired group vs. 17 years in conventional adult group, P < 0.0001) no statistically significant heterogeneity was demonstrated.
This systematic review and meta-analysis of 11 published comparative studies, reporting results of 2,489 AYA patients with ALL showed a significant reduction in all cause mortality in patients given pediatric-inspired regimens compared with patients given conventional adult regimens (RR 0.59; 95% CI 0.52–0.66). This was confirmed in a sensitivity analysis including only studies in which the two cohorts were well compared in either age or disease risk. Post-induction CR rate and EFS increased with pediatric-inspired regimens, and the relapse rate decreased. Of note, there was no difference in NRM between the two groups.
Our systematic review also compared for each drug, the cumulative doses administered in pediatric regimens compared with conventional adult chemotherapy. It demonstrated that relatively higher doses of cytostatic agents (corticosteroids, vincristine, and PEG-asparaginase along with early and repeated CNS prophylaxis) and lower doses of cytotoxic medications (daunorubicin, cytarabine, and etoposide) are used in the pediatric-inspired regimens as compared with the conventional adult protocols. However, this comparison evaluates only intention to treat drugs' cumulative doses and not their dose intensity. Based on the current data, few expert opinion guidelines suggested that the best therapeutic approach for an AYA patient would be an intensive pediatric regimen [2–4,31] followed by an extended maintenance therapy. However, as shown in this review, these data are based on nonrandomized comparative studies, which are prone to bias and data from randomized controlled trials are currently not available. Furthermore, for patients older than 20 years, these conclusions are questionable. The two studies that included patients older than 20 years did not perform a subgroup analysis and it is not clear if patients older than 20 years benefit from pediatric-inspired regimens.
A number of important differences between younger and older patients with ALL exist. These potentially can lead to bias and should be taken into account when interpreting results of studies, which are not randomized controlled trials.
First, the biology of both the underlying disease and the patients' changes with age and when comparing outcomes between two cohorts, patients' age should be well matched [32–37]. In this systematic review disease risk or age were well matched between the two cohorts in only in five out of the 11 included studies.
The second major difference between children, AYA, adult and elderly patients is the difference in therapy-related toxicity [38, 39]. Thus, if pediatric-inspired regimens are expected to be more intensive and toxic than the standard regimens, an increase in NRM is also likely. However, NRM in the present meta-analysis was similar for both groups. This result should be interpreted cautiously as the majority of patients included in these trials were younger than 20 years and several groups applying pediatric-inspired protocols to adults reported a significant increase in treatment-associated toxicity in the older population of the AYA group [17, 39].
The third potential cause for a superior outcome in the younger population relates to the protocols administered: as stated above, typical treatment protocols for adults may not be intensive enough for AYA; yet, they are too intensive for older or elderly patients . In our systematic review, we showed that most pediatric-inspired regimens used higher doses of corticosteroids, vincristine, asparaginase, and more frequent CNS prophylaxis. Importantly, the Finnish group that used comparable doses in both the pediatric-inspired and the standard adult regimens, as opposed to other studies in our meta-analysis, reported similar CR and all cause mortality rates for both arms .
One should be cautious when translating results of the meta-analysis to clinical practice. The main drawback of our meta-analysis is the fact that none of the studies were randomized. Second, conclusions are largely applicable for adolescents up to the age of 20 years but it is not clear whether pediatric-inspired regimens are the best treatment strategy for AYA patients older than 20.
It is difficult to define which component the treatment of AYA is responsible for the apparently improved outcome with pediatric-inspired protocol/regimen: the treating physician (pediatrician vs. adult oncologist), patients' compliance and social support or merely the prescribed regimen. Finally, the role of minimal residual disease and its implementation has not been assessed yet despite recent findings regarding its major prognostic value and its role in treatment stratification and adjustment .
Implications for practice and for research
The complexity of ALL treatment protocols and the importance of adherence to therapy require AYAs to be treated in centers with much experience in this disease. If possible, AYAs and older adults should be enrolled into randomized controlled trials further investigating pediatric-like approaches versus adult protocols. Setting the age limit for the feasibility of these protocols is crucial.
We thank the authors who responded to letters and supplied additional data on their trials. We also thank Prof. Baruch Wolach for assistance in translation of non-English published articles.