Introduction
The first successful transplants of hematopoietic stem cells were performed in 1968 in three children with congenital immune deficiency diseases.[1-4] In each instance, stem cells were collected from the bone marrow of sibling donors who were genotypically identical or closely matched to the recipient for human leukocyte antigens (HLA). Since then, thousands of patients have received hematopoietic stem cell transplants as treatment for malignant and nonmalignant diseases.
Approximately 50,000 transplants are performed worldwide each year (Figure 1). Reasons for the increased use of stem cell transplantation over the past 3 decades include its proven and potential efficacy in many diseases, better understanding of the appropriate timing of transplantation and patient selection, greater availability of donors, better techniques for determining HLA match, greater ease of stem cell collection, and improved supportive care resulting in less transplant-related morbidity and mortality.
About two-thirds of hematopoietic stem cell transplants use autologous cells, generally extracted from peripheral blood by leukapheresis. The remainder are allogeneic transplants, most commonly using cells collected directly from the bone marrow of HLA-identical sibling donors (Figure 1 and Figure 2).
The growth of hematopoietic stem cell transplantation has been accompanied by a coordinated, international effort to collect and analyze data on transplant outcomes through the International Bone Marrow Transplant Registry (IBMTR), established in 1972, and the Autologous Blood and Marrow Transplant Registry (ABMTR), established in 1990. Over 350 institutions in 47 countries contribute data to the IBMTR, and over 250 institutions in North and South America contribute data to the ABMTR. Participating centers submit data on their consecutive transplants to the IBMTR/ABMTR Statistical Center. The Center receives data on more than 12,000 new transplants each year and maintains a database that now includes information on more than 120,000 transplant recipients.
Organizational Structure of the IBMTR/ABMTR
The activities of the IBMTR and ABMTR are supervised by Advisory, Executive, and Working Committees and a joint IBMTR/ABMTR Steering Committee (Figure 3). The Advisory and Executive Committees review policies for the use of IBMTR/ABMTR data and advise the scientific and statistical directors on administrative and scientific matters. The Working Committees design and conduct studies that are relevant to their subject area, consider proposals to use IBMTR/ABMTR data for specific studies, periodically assess and revise relevant sections of IBMTR/ABMTR data collection forms, and plan and conduct workshops at IBMTR/ABMTR meetings. The Steering Committee establishes priorities for scientific activities.
Since 1972, the IBMTR/ABMTR Statistical Center has been central to Registry activities, coordinating data collection and management, and providing statistical and administrative support for studies using Registry data. The Statistical Center is an academic division of the Health Policy Institute of the Medical College of Wisconsin in Milwaukee.
Data Collection
The IBMTR/ABMTR collects data on two levels: registration and research. Registration data include disease type, age, sex, pretransplant disease stage and response to chemotherapy, date of diagnosis, donor type, graft type (bone marrow- and/or blood-derived stem cells), transplant regimen, posttransplant disease progression and survival, engraftment, graft-vs-host disease (GVHD), development of a new malignancy, and cause of death. All IBMTR/ABMTR centers contribute registration data.
Research data are submitted on comprehensive report forms completed for a subset of registered patients in IBMTR/ABMTR research centers. Research data include detailed pre- and posttransplant clinical information such as disease subtype, tumor size and pathology, sites of disease, nontransplant treatment of the primary disease, performance status, organ function, details of the transplant regimen including dose and schedule of high-dose therapy, graft manipulation, supportive care, posttransplant toxicities, and functional status.
Both databases are longitudinal; patients are followed through their transplant centers with yearly updates.
The Use of Multicenter Observational Databases
Transplant outcomes are influenced by many patient- and disease-related factors (such as age, disease stage and prior treatment), as well as transplant-related factors such as stem cell source, conditioning regimen, and prophylaxis for GVHD. Ideally, most transplant strategies would be evaluated by large randomized clinical trials. However, various factors limit the application of randomized trials in hematopoietic stem cell transplantation. Many diseases treated with transplants are uncommon; thus, single centers may treat only a few patients with a given disorder. This makes randomized trials difficult and also limits the ability to perform nonrandomized (phase II) trials with sufficient power to detect meaningful effects. Small trials, even when randomized, may provide misleading results.[5]
New transplant technologies are rapidly being introduced, so the results of prospective clinical trials may be obsolete before they are published. Some important transplant issues are not amenable to randomization, eg, differences in outcome associated with differences in donor type. In a systematic review of 255 transplant-related studies published between 1990 and 1992, only 16 (6%) were randomized trials; most of these studies had fewer than 100 patients.[6]
Even when randomized trials are performed, enrolled patients may represent only a small proportion of the target population and may not be representative of the larger group.[7-10] The results of treatments administered in these trials may differ from those obtained when the technology is more widely applied. Most clinical trials focus on short- and intermediate-term outcomes (1 to 5 years). However, there is a need for long-term follow-up of transplant recipients because high-dose therapy may be associated with important effects, such as therapy-related cancers, that may not develop until years after the transplant was performed.
Observational databases may facilitate our understanding of transplant outcomes by addressing questions that are difficult to address in randomized trials. These include descriptions of transplant results in various disease states and patient groups; analysis of prognostic factors; evaluation of new transplant regimens; comparison of transplant with nontransplant therapy; defining intercenter variability in diagnosis, practice, and outcome; and developing analytic approaches to evaluating transplantation outcomes and costs.
The value of observational studies in assessing treatment effects was highlighted by two recent articles published in the New England Journal of Medicine.[11,12] Clinical databases may also be useful in developing optimal designs for randomized studies and in interpreting the results of such studies.
Descriptive Studies
Databases that include a large number of centers involved in allogeneic and autologous transplantation are uniquely suited for descriptive studies. This is particularly important in the case of rare diseases or more common conditions for which transplantation is infrequently performed. In these situations, single centers often have only one or a few cases, precluding meaningful assessment of outcome. The publication of outcomes may be biased to results that are particularly good or bad.[13] By combining data from many centers and obtaining data systematically on all transplants, regardless of outcome, registries can provide a more precise and unbiased estimate of results. Examples of descriptive studies using IBMTR/ABMTR data are analyses of transplants for Ph+ acute lymphoblastic leukemia,[14] Diamond-Blackfan anemia,[15] chronic lymphocytic leukemia,[16] and paroxysmal nocturnal hemoglobinuria.[17]
Identification of Prognostic Factors
The heterogeneity and large numbers of patients reported to the IBMTR/ABMTR allow use of multivariate regression techniques to evaluate associations between patient- and disease-related variables and outcome. Because transplant centers must report all consecutive transplant recipients, the full range of characteristics found in transplant patients is available for study. This is an important use of large observational databases, since many prognostic factor studies are limited by small numbers, nonrepresentative populations, and/or insufficient detail on patient and disease characteristics.[18-21]
Examples of large prognostic factor studies using IBMTR/ABMTR data include assessment of (1) risk factors for acute and chronic GVHD,[22,23] interstitial pneumonia,[24,25] and veno-occlusive disease of the liver[26]; (2) prognostic factors for relapse and leukemia-free survival after transplants for acute myelogenous leukemia, acute lymphoblastic leukemia, and chronic myelogenous leukemia[27-30]; (3) risk factors for graft failure after transplants for severe aplastic anemia[31]; and (4) factors associated with the outcome of autotransplants for metastatic breast cancer.[32]
