First-Line Matched Related Donor Hematopoietic Stem Cell Transplantation Compared to Immunosuppressive Therapy in Acquired Severe Aplastic Anemia

Introduction Acquired severe aplastic anemia (SAA) is a rare and progressive disease characterized by an immune-mediated functional impairment of hematopoietic stem cells. Transplantation of these cells is a first-line treatment option if HLA-matched related donors are available. First-line immunosuppressive therapy may be offered as alternative. The aim was to compare the outcome of these patients in controlled trials. Methods A systematic search was performed in the bibliographic databases MEDLINE, EMBASE, and The Cochrane Library. To show an overview of various outcomes by treatment group we conducted a meta-analysis on overall survival. We evaluated whether studies reported statistically significant factors for improved survival. Results 26 non-randomized controlled trials (7,955 patients enrolled from 1970 to 2001) were identified. We did not identify any RCTs. Risk of bias was high except in 4 studies. Young age and recent year of treatment were identified as factors for improved survival in the HSCT group. Advanced age, SAA without very severe aplastic anemia, and combination of anti-lymphocyte globulin with cyclosporine A were factors for improved survival in the IST group. In 19 studies (4,855 patients), summary statistics were sufficient to be included in meta-analysis. Considerable heterogeneity did not justify a pooled estimate. Adverse events were inconsistently reported and varied significantly across studies. Conclusions Young age and recent year of treatment were identified as factors for improved survival in the transplant group. Advanced age, SAA without very severe aplastic anemia, and combination of anti-lymphocyte globulin with cyclosporine A were factors for improved survival in the immunosuppressive group. Considerable heterogeneity of non-randomized controlled studies did not justify a pooled estimate. Adverse events were inconsistently reported and varied significantly across studies.


Introduction
Acquired severe aplastic anemia (SAA) is a rare [1] and potentially fatal disease which is characterized by hypocellular bone marrow and pancytopenia, and mainly affects young adults. The incidence rate was estimated at less than 4 per million people per year [2]. The major signs and symptoms are severe infections, bleeding, and exhaustion. The underlying pathophysiology is thought to be an aberrant immune response involving the T-cell mediated destruction of hematopoietic stem cells. In most cases, the cause is unknown, although various triggers such as drugs, toxins, and viruses have been reported [3,4].
The treatment of SAA mainly includes immunosuppressive therapy (IST) with antithymocyte globulin (ATG)/antilymphocyte globulin (ALG) and cyclosporine A (CSA), or allogeneic hematopoietic stem cell transplantation (HSCT) [3,4,5]. Allogeneic HSCT is seen as the treatment of choice for selected patients with an HLA-matched related donor [6,7]. Allogeneic HSCT is associated with graft failure, graft-versus-host disease (GVHD), and organ toxicities. On the other hand, patients may not respond to IST and long-term IST is associated with the development of clonal diseases [8]. Clinical treatment algorithms have been suggested to find a decision that meets individual conditions, personal preferences, and prognostic factors [9].
The present systematic review and meta-analysis compares the outcome after first-line HLA-matched related donor HSCT vs. IST in SAA patients in published controlled trials.

Methods
While preparing this systematic review and meta-analysis, we endorsed the PRISMA statement, adhered to its principles and conformed to its checklist [10,11].

Study inclusion criteria
We included patients with acquired severe aplastic anemia who received, as first-line treatment, allogeneic HSCT from HLA-matched related donors (MRD) as the test intervention and IST as the control intervention. Study design was limited to randomized controlled trials (RCTs) and non-randomized intervention studies [12]. We did not set a minimum sample size to be considered. Fulltext publications in English language were considered. We set no limits on year of publication or year of treatment. A protocol is not available.
Search strategy MEDLINE (1950MEDLINE ( to 2010, EMBASE (1980EMBASE ( to 2010 and The Cochrane Library (to 2010) were searched without restrictions on study design and publication year (final search 10 January 2010). The first database search was conducted 24 January 2006. The final search 10 January 2010 included a modified strategy to consider MeSH changes and to render the exclusion of animal studies more precisely. The MeSH term BONE MARROW TRANSPLANTATION was deleted from one category [13]. For MeSH 2008 there was a major revision of Publication Types (PT) and the phrase ''as Topic'' was added [14]. We introduced (ANIMALS not (ANIMALS and HUMANS)).sh. and replaced (ANIMALS not HUMANS).sh. To account for these changes, both searches were not restricted to any publication year. The terms and the syntax used for the search in MEDLINE via Ovid as shown in Table 1 were tailored to the requirements of the other 2 databases. Reference lists of all included original articles and 5 recent reviews (2007 to 2009) [8,9,15,16,17] were handsearched. Abstracts of the American Society of Hematology Annual Meeting 2004 to 2009 [18] and information on studies registered at ClinicalTrials.gov [19] were searched online (April 2010).

Study selection
First, articles were excluded if the title and/or the abstract clearly referred to other diagnoses than severe aplastic anemia and in addition clearly referred to other interventions than allogeneic hematopoietic stem cell transplantation. Second, articles not excluded in the first step were evaluated whether patients were analyzed in a test group after first-line allogeneic HSCT from a MRD and were compared with patients after first-line IST in a control group. Reporting of extractable information about overall survival was required for all included studies. For each excluded study, an appropriate reason was documented ( Figure 1). All steps of the literature screening process were performed by two independent reviewers. Any disagreements were resolved by discussion. Criteria for classification of severity of aplastic anemia and quality of response after immunosuppressive treatment were applied according to the EBMT [21] [20].

Risk of bias
Risk of bias within studies was evaluated by assessing study design, such as retro-or prospective planning, concurrent control group, criteria for assignment of patients to treatment arms, control for confounding factors, and other criteria, such as unclear selection of patients and analysis of the same patients in both treatment groups, that may increase the risk of bias especially in non-randomized trials [21]. A low risk of bias required a 'yes' for all three of the following topics: concurrent control group, control for confounding factors, and no other risk of bias factors.
Risk of bias across studies was evaluated by assessing publication bias and outcome reporting bias. We conducted funnel plots using hazard ratios and related standard errors of each meta-analysis to assess publication bias. We evaluated potentially relevant studies to identify studies that may have been excluded because of missing or insufficient outcome reporting. We evaluated published study protocols to identify outcome reporting different from appropriate procotols.

Primary outcome: overall survival
The primary effect measure for meta-analyses was the hazard ratio. If the hazard ratio was not directly given in the publication, we extracted summary statistics from Kaplan Meier survival functions and estimated hazard ratios according to methods proposed by Parmar 1998 [22]. For estimation, we applied a tool which uses p-values of the appropriate log-rank test comparing the two survival functions of interest, number of patients analyzed, and number of events on each arm [23]. If this information was not available, hazard ratios were deduced from the graphical display of the survival curves, if possible. Metaanalyses were conducted using the generic variance approach [24,25] and the random effects model [26]. Calculations were conducted using SAS version 9.2 (SAS Institute Inc., Cary, North Carolina, USA). The results of the meta-analyses were graphically displayed by means of a forest plot. Heterogeneity of the results was visually assessed and quantified using the I 2 value [27]. A funnel plot was prepared by using The Cochrane Collaboration's Review Manager 5 (http://www.cc-ims.net/ revman). In the case of considerable heterogeneity (I 2 $50%), a pooled estimate is not reasonable and, therefore, was not calculated [28].

Subgroup analysis
Sensitivity analysis: We tried to explain heterogeneity identified in the meta-analysis of overall survival by evaluating dichotomized subpopulations of study and patients characteristics in several sensitivity analyses [29]. Subgroups in individual studies: We extracted overall survival data of subgroups to reproduce survival functions from some individual studies if reported for both treatments. Data on subgroup analyses were extracted if available. The difference between the estimates of 2 subgroups was evaluated by conducting a test of interaction based on meta-regression models [30]. Calculations were conducted using SAS version 9.2 (SAS Institute Inc., Cary, North Carolina, USA).

Factors for improved survival
We extracted statistically significant results from multivariate analyses on factors for improved survival, such as age, disease severity, treatment period, and type of IST, if reported in the included studies.

Data collection and analysis
All steps of the data collection process were performed by two independent reviewers. Any disagreements were resolved by discussion. We collected study characteristics such as the number and region of participating centers, the treatment period, the number of analyzed patients per treatment arm, the proportion of patients with diseases other then SAA, a regimen other than firstline treatment, and median follow-up. Median age, gender, and time interval from diagnosis to treatment were extracted as patients' characteristics. For the HSCT arm we extracted the number of patients who were treated with conditioning irradiation and conditioning ATG or ALG. For the IST arm we extracted the number of patients treated with ATG or ALG alone or in combination with CSA.
The primary outcome was overall survival from the beginning of HSCT or IST. Kaplan Meier estimates were extracted directly from the text or deduced from survival curves given in the publication. The principal summary measure was the hazard ratio as specified in the meta-analysis section. We extracted the p-value of the log-rank test of the overall survival functions and the 5-year estimate of both treatment arms. Results from subgroup analyses were also extracted.
The secondary outcomes were overall mortality, graft failure, GVHD, no response to IST, and health-related quality of life.  First-Line HSCT Compared to IST in SAA PLoS ONE | www.plosone.org Graft failure included both primary and secondary types; acute GVHD was considered if grade II-IV was reported and chronic GVHD was considered if an extensive course was reported.

Baseline data
An overview of characteristics of 7,955 analyzed patients in 26 studies is presented in Table 3 An overview of the treatment characteristics is presented in Table 4. Conditioning irradiation was reported in 22 of 26 studies (0%-100%) and ATG or ALG was used in 9 studies (2%-100%) in the HSCT arms. In the IST arms, ATG or ALG (8%-100%) and CSA (0%-100%) was reported in 25 of 26 studies. IST was composed of ATG or ALG in 25 studies and was combined with CSA in 13 studies.

Risk of bias
Risk of bias within studies was high for 21 of 26 included studies mainly due to flaws of study design, assignment of patients to treatment groups, and missing control for confounding ( Table 6). In 7 studies, an upper age limit was applied as additional assignment criteria.
The funnel plot of 19 studies included in meta-analysis shows no asymmetry (Figure 3), which may be consistent with absent publication bias. In 7 of 26 included studies, data of the primary outcome were not sufficiently reported to be included in the main meta-analysis.

Subgroup and sensitivity analysis
The difference between the estimates was not statistically significant in both approaches and for all evaluated items, except *other than SAA: MAA or not reported degree of severity.    for age, which was statistically significant in the evaluation of subgroups from individual studies and not statistically significant in the sensitivity analysis (Table 7).

Factors for improved overall survival
We evaluated patient characteristics that were evaluated in multivariate regression analyses (Table 8). 11 studies reported statistically significant factors for improved survival and 15 studies did not. Frequency of reporting the main factors, that is, age, year of transplant and type of IST is presented in Table 9.
Young age was identified as a statistically significant factor for improved overall survival for patients in the HSCT group in 5 studies. The results for patients in the IST group were less clear because advanced age was a favorable factor in 3 studies and young age a favorable factor in 2 studies. The actual age limit, which was used to dichotomize participants into young vs. advanced age varied across studies. Level of polymorphonuclear neutrophilic granulocytes (PMN) between 0.2 G/L and 0.5 G/L was found to be an important factor for improved survival after IST in 5 studies. Only 1 study found an advantage with less than 0.2 G/L in patients who were younger than 17 years. Recent year of treatment was an important factor for improved survival after HSCT in 4 studies and after IST in 1 study. Improvement of outcome over time was accompanied with treatment refinement, such as combination of methotrexate with CSA for GVHD prophylaxis and reduction of conditioning irradiation. Another example of improvement of outcome with treatment refinement is the combination of ALG and CSA instead of monotherapy of ALG. This combination was identified as factor for improved survival in 2 studies.

Secondary outcomes
An overview of graft failure, GVHD, no response, and overall mortality is presented in Table 10. Graft failure (reported in 15 studies) ranged from 1% to 43%, acute GVHD grade II to IV (reported in 16 studies) ranged from 0% to 88%, and extensive   (68)

Primary outcome
We found high risk of bias among 26 identified nonrandomized controlled studies. Considerable heterogeneity of 19 studies included in a meta-analysis of overall survival did not justify a pooled estimate. The aim to compare the primary outcome between treatment groups in an overall synthesis of available data was not achieved. We were able to identify statistically significant factors for improved overall survival reported in the studies. Young age (rather than advanced age) and recent (rather than earlier) year of treatment were associated with a better overall survival in the HSCT group. Advanced age (rather than young age), SAA without VSAA (rather than VSAA), and combination of ALG and CSA (rather than ALG alone) were associated with a better overall survival in the IST group. While pooling data on overall survival of all participating patients did not appear sensible, we conducted sensitivity meta-analysis of subgroups to explain heterogeneity and of subgroups reported in individual studies. Unfortunately, these evaluations were hampered by the fact that individual patient data were not available and instead of that we relied on published aggregate data if reported. It should be mentioned that ALG is no longer available.
Although young age was recognized as a major influence factor on overall survival, the appropriate definition for young age varies considerably across studies. For example, the cut-off for young vs. advanced age was 16 years in the study of Bacigalupo 2008 [16], 30 years in the study of Ljungman 2009 [59], and 40 years in the study of Marsh 2009 [9].
Bacigalupo 2008 [16] reported that the outcome has improved since 1996 for HSCT but not for IST. This result was supported by 4 studies whereas 1 study did find an improvement for the IST group as well. Several factors may have contributed to recent improvements for HSCT, such as detailed HLA-matching, less irradiation-based conditioning, less acute GVHD with a prophylaxis combination of methotrexate plus CSA instead of methotrexate alone [34], refinement of the type and dosage of conditioning drugs, and general advancement of medical and nursing clinical science.
We found that a combination therapy of ALG plus CSA favored overall survival in the IST group, underscoring that refinement of therapy obviously has also improved the outcome with IST. Gafter-Gvili 2008 [60] concluded in a systematic review and metaanalysis that combination of ALG plus CSA should be considered the gold standard for IST for patients with SAA.
6 studies [33,37,38,39,41,50] evaluated whether the disease severity of VSAA (PMN ,0.2 G/L) vs. SAA (without VSAA; PMN 0.2 to ,0.5 G/L) had an impact on overall survival. 5 studies [33,37,38,39,50] consistently found that SAA without VSAA favored the outcome in IST group when compared to VSAA. 2 studies [39,50] included only children and 3 studies [33,37,38] included also adults. These results may suggest that HSCT may be the preferred treatment option for patients with VSAA and that IST may be the preferred option for patients with SAA without VSAA. However, Führer 2005 [41] reported contradictory results that VSAA favored the outcome in IST group. The results may be relevant for children only because all analyzed patients were younger than 17 years of age. We did not find another study confirming the results, especially not in 2 studies mentioned above.
We believe that this is the first comprehensive systematic review and meta-analysis about studies comparing first-line HSCT versus first-line IST in patients with SAA. First-Line HSCT Compared to IST in SAA PLoS ONE | www.plosone.org

Secondary outcomes
Studies inconsistently reported adverse events and their frequencies varied significantly across studies. Graft failure was highest in early studies but could affect up to 18% of patients in recent studies. Rates for acute GVHD grade III to IV reached up to 40% and for extensive chronic GVHD reached up to 24% in recent studies. A considerable number of patients not responding to IST may indicate the importance of unrelated donor transplantation.

Duplicate publication bias
Identical data may have been included twice in the present systematic review. We searched for follow-up papers of a single study to include the update version and exclude former presentations. Register analyses can provide results based on a large number of patients but registers may use data that may have been published previously by the contributing study centers. We identified 5 studies published on behalf of the EBMT [33,34,50,51] which are probably based on overlapping data. Locasciulli 2007 [51] presented an update of the EBMT data and, therefore, investigated considerably more patients (2479 patients) than Locasciulli 1990 [50] (304 patients). The courses of the survival functions are clearly different between the 2 studies. In Locasciulli 1990 [50] (Figure 1 of the article), in the majority of the follow-up period the course of the Kaplan-Meier curve after HSCT is above that after IST (congruent in the first 12 months), indicating an advantage from HSCT. In Locasciulli 2007 [51] ( Figure 1 of the article), in the majority of the follow-up period the course of the Kaplan-Meier curve after HSCT is below that after IST (in the first 60 months). Only at the very end of the follow-up period (from 100 to 120 months) is the HSCT curve above the IST curve. The difference between the 2 survival functions was assessed by a log-rank test. The authors stated that 10-year survival was significantly superior in patients treated with BMT than in those in whom immunosuppression was used (73% versus 68%, p = 0.002). On the contrary, we are convinced that the survival functions clearly show that IST was statistically significantly better than HSCT. Our

Outcome reporting bias
Outcome reporting bias [62] is defined as the selection of a subset of the originally recorded outcome variables for publication. Systematic reviews need to address the issue of missing outcome data because outcome reporting bias can affect their conclusions [63]. We identified considerable outcome reporting bias. A major flaw of all studies was the lack of statistical summary data such as standard

Study publication bias
Study publication bias is defined as publication of research results depending on their results [64]. Funnel plots of both reported meta-analyses show moderate asymmetry and do not indicate considerable publication bias. The strengths of the present systematic review are the broadness of the search strategy and the comprehensiveness of the published data included. Nevertheless, there may be a slight possibility that an unknown number of studies were not registered and not published.

Language bias
Results in English language articles could be different from those of articles written in other languages [65]. Non-English language articles require expensive translations to prevent selective outcome extraction and misinterpretation of results. Funding for translation was not provided and we excluded all non-English language articles, including German articles. Restricting the inclusion of studies to English articles may have little effect on summary treatment effect estimates [66,67] and German language articles may not play a preeminent role in the dissemination of medical research [68].

Internal validity
We identified a high risk of bias within all non-randomized controlled studies except for 4 studies. Assignment of patients to treatment groups was reported to be based on availability of MRD in all 26 studies, although, in 7 studies an upper age limit was -: information extractable from publication. *Bayever 1984: acute GVHD including interstitial pneumonia; chronic GVHD including moderate or severe types. Abbreviations: GVHD: graft-versus-host disease; HSCT: hematopoietic stem cell transplantation; IST: immunosuppressive therapy.

{
Bacigalupo 2000: less acute GVHD with methotrexate plus cyclosporine A vs. methotrexate or cyclosporine A alone; incidence of acute GVHD III to IV has been reduced from 20% to 6% with time (p,0.00001), however, survial for patients with acute GVHD.I has not improved with time; incidence of extensive chronic GVHD has been reduced from 15% to 5% with time (p,0.00001), however, survival for patients with extensive chronic GVHD beyond day 100 after HSCT has not been improved with time. Abbreviations: GVHD: graft-versus-host disease; HSCT: hematopoietic stem cell transplantation; IST: immunosuppressive therapy. doi:10.1371/journal.pone.0018572.t010 applied as additional assignment criteria. This type of allocation has specific requirements, such as allelic vs. serologic typing, number of analyzed loci, time spent searching for donors, documentation of all individuals analyzed including families of IST patients, number of analyzed individuals per family, intent-totreat analysis. However, these requirements were not reported in the included studies.

Heterogeneity
We included as many controlled studies as possible in order to not miss any valuable outcome information. Consequently, we accepted studies that considered a considerable number of patients with moderate aplastic anemia or unknown disease severity. Pooling of data from patients with varying characteristics may have compromised the generalizability of results. Furthermore, important subgroups such as young and advanced age were confused in the meta-analysis.
In an attempt to reduce heterogeneity we strictly confined the meta-analysis to studies that clearly included at least 80% of patients with SAA and first-line treatment (data not shown). We found a moderate heterogeneity and a statistically significant pooled estimate that favored HSCT. Many studies including large and recently published ones were excluded. The result suggested a global preference for one treatment disregarding conditions other than the treatment that might have a determined impact on the outcome. We believe that this procedure would have introduced a study selection bias and a misleading conclusion and was therefore not pursued.

Strengths and limitations of the present review
The strengths of this review are the broadness of the search strategy and the comprehensiveness of the published data included. Significant factors that may influence the survival of the patients were considered in the present systematic review. We estimated hazard ratios from published aggregate survival functions and did not use individual patient data. Subgroup analysis was not helpful to explain considerable heterogeneity found in meta-analysis. While the results of the meta-analysis may not be conclusive, they can provide useful summaries of the state of knowledge.

Conclusions
Young age and recent year of treatment were identified as factors for improved survival in the transplant group. Advanced age, SAA without very severe aplastic anemia, and combination of anti-lymphocyte globulin with cyclosporine A were factors for improved survival in the immunosuppressive group. Considerable heterogeneity of non-randomized controlled studies did not justify a pooled estimate. Adverse events were inconsistently reported and varied significantly across studies.