Introduction

Hematopoietic stem cell transplantation (HCT) is performed to treat various malignant and non-malignant hematological diseases. Transplant outcome has improved over the past decades, and resulted in increasing numbers of long-term transplant survivors [1,2,3,4,5,6,7]. Although many transplant survivors live healthy lives, the delayed effects of transplantation as well as increased risk of secondary cancers continue to affect the clinical courses of transplant survivors [8].

In the field of cancer studies, population-based descriptive studies are a commonly used methodology [9,10,11,12]. Cancer patients may die from causes other than the studied cancer; therefore, the methodology used in population-based descriptive studies removes risks of dying from other causes from the assessment and estimates the proportion of patients dying from direct or indirect consequences of cancer [13, 14]. Transplant recipients may die of the causes other than original disease, which are commonly grouped as non-relapse mortality. It is assumed that a significant portion of these deaths are associated with the transplantation process. However, this assumption may not be valid for individuals who survive for an extended period, as they could potentially face mortality from causes unrelated to the transplant [15]. Few population-based studies have been performed in transplant settings [7, 16,17,18]; however, population-based descriptive epidemiology studies employing methodologies such as net survival may also provide useful information for patients receiving transplants similarly to cancer survival studies while excluding the risks of dying from causes other than cancer or transplantation.

Previous studies have investigated the prognosis of patients who would receive a future transplant. Statistical time analyses in the majority of previous transplant outcome studies are generally performed using the time of transplantation as the time origin [19]. However, information on survival trends obtained from those studies are limited for patients who have already survived more than a couple of years post-transplant [20]. Conditional survival (CS) is defined as the probability that a person will survive an additional number of years if they have already survived a specific number of years. This metric offers an updated prediction of survival throughout the course of the disease [21, 22]. For long-term transplant survivors, CS may provide more relevant information compared to survival measured from the time of transplantation. Meanwhile, the graft source differently affects the course of the transplant and survivals due to the factors such as delayed engraftment, risk of infection, acute and chronic graft-versus-host disease (GvHD). Although many studies have reported differences in transplant outcomes between bone marrow, peripheral blood (PB), and cord blood (CB) for various diseases [23,24,25,26], these studies were generally analyzed from the time of transplant. Information of long-term survival probabilities after the 2-year period when most deaths occur are scarce.

This study aimed to evaluate the outcomes of transplant survivors from year 0 to year 5 after transplant. We also aimed to provide an overview of long-term survival of children and adults who had already survived up to 5 years from the time of transplant, by investigating different graft sources for different diseases using a population-based methodology incorporating net-survival analysis [27] and CS using net-survival probabilities.

Methods

Patients and data

Data were obtained from the Japanese Data Center for Hematopoietic Cell Transplantation (JDCHCT) [28,29,30,31,32]. Survival information for patients were annually updated at each transplant center, and the final annual update cycle of this dataset was performed during the period from June to October, 2016.

Patients included in the analysis were patients who received first allogeneic HCT between 1991 and 2015. Diseases consisted of: acute myeloid leukemia (AML), acute lymphoblastic leukemia (ALL), adult T-cell leukemia/lymphoma (ATL), chronic myeloid leukemia (CML), myelodysplastic syndrome (MDS), malignant lymphoma (ML), multiple myeloma (MM), and aplastic anemia (AA). For age group analysis, children were defined as those 15 years old and under, younger adults as those aged 16 years to 59 years old, and older adults as those aged 60 years and over. Patients with unknown sex were excluded. In total, 41,716 patients were eligible. The study was approved by Nagoya University Ethics Committee (approval no. 2018-0150) and was conducted in accordance with the Declaration of Helsinki. The patients provided informed consent for the use of clinical information.

Outcomes and statistical analysis

The main outcomes of the study were 10-year net survival and the conditional 5-year net survival at 5 years after transplantation, which is defined as the net probability of surviving a further 5 years given that the patient had already survived for 5 years. In cancer studies, net survival provides the cumulative probability for cancer patients to survive their cancer up to an elapsed point in time from the starting time, such as the time of diagnosis or time of transplantation, after controlling for risks of death from other background mortalities (see Supplementary Materials for further details) [10, 13, 14, 33]. Survival was defined as the time after transplantation or landmark time until death from any cause, with surviving patients censored at the time of last follow up. Conditional 5-year net survival by disease and age groups were calculated for patients who had already survived 5 years after transplantation. The Pohar-Perme method was applied for the estimation of net survival. The background mortality of the patients was calculated using the complete Japanese population life tables by birth year, age, and sex. These tables are published annually by Center for Cancer Control and Information Services, National Cancer Center, which is based on information provided by the Statistics and Information Department of the Ministry of Health, Labour and Welfare, Japan [34, 35].

For each disease, graft source, and age group sub-analysis, conditional 5-year net survival was further estimated annually for patients who had survived up to 5 years after the transplant. Upper and lower boundaries of 95% confidence interval (CI) were used to compare the survivals of each graft source. For younger adults, AML, ALL, ATL, CML, MDS, MM, ML, and AA were analyzed; however, we only analyzed AML, ALL, MDS, ML, and AA for children, and AML, ALL, MDS, ML, and ATL for older adults, due to the limited number of patients with other diseases. Overall survival (OS) and conditional OS were estimated using Kaplan–Meier method (as described in the Supplementary Materials). Causes of death and completeness of follow-up [36, 37] are also summarized in the Supplementary materials. Stata version 17.0 (StataCorp, TX) was used for analyses.

Results

Characteristics of patients and transplantation outcomes

The characteristics of the patients, their diseases, and transplants received are shown in Table 1. The median age of the recipients was 41 years. AML (39%), ALL (22%), MDS (11%), and ML (10%) were the most frequent indications for transplantation. Bone marrow transplantation accounted for 60%, PB accounted for 19%, and CB was used in 20%. Detailed patient characteristics according to graft source and age groups are shown in Table S1. Completeness of follow-up at 5 years after transplantation for patients who received a transplantation between 1991 and May 2011 was 92.9%.

Table 1 Patient characteristics.
Full size table

Net survival and conditional net survival for each disease

We calculated net survival by sex and disease. Five-year net survivals were 45.5% (95% CI, 44.9–46.2) and 50.8% (50.0–51.6), and 10-year net survivals were 41.5% (40.7–42.2) and 47.4% (46.6–48.3), for males and females respectively (Fig. 1a). Ten-year net survival estimates for each disease were relatively high for AA (78.1% [76.2–80.1]), and relatively low for ATL (27.6% [24.9–30.4]) and MM (21.1% [15.6–26.6]) (Table 2 and Fig. 1b). Net survival of CML, ALL, MDS, ML, and AML ranged intermediately.

Fig. 1: Net survival estimates of overall patients by sex and disease.
figure 1

Net survival of the overall patients by (a) sex and (b) disease.

Full size image
Table 2 Ten-year net survival and 5-year conditional net survival of 5-year survivors.
Full size table

Five-year CS at 5 years after transplantation by disease and age group is shown in Table 2. For the all-age and younger adult cohorts, 5-year CS of 5-year transplant survivors was relatively low for MM compared to other diseases (77.9% [95% CI, 65.5–90.3] and 76.6% [63.4–89.8], respectively). For children, 5-year CS of 5-year transplant survivors with AML, ALL, MDS, ML, and AA were relatively high with point estimates above 90%.

Conditional net survival of each graft source

Five-year conditional net survival by disease and age categories is shown in Figs. 2 and 3. Corresponding point estimates and 95% CIs are shown in Table S2. The impact of age on each disease is also illustrated for AML, ALL, MDS, ML, AA, and ATL by graft source and diseases in Fig. S1, using the same point estimates.

Fig. 2: Five-year conditional net survivals of the all-patient cohort and younger adults who have survived up to 5 years after transplantation.
figure 2

Five-year conditional net survivals of: (a) all ages and (b) younger adult patients at year 0 to year 5 after the transplant. Five-year conditional survival CS (t = 5 | X) was defined as the probability of surviving an additional number of years (t = 5), conditional on survival of (X) years where X = 0, 1, 2, 3, 4 and 5. For each disease, changes of 5-year conditional net survivals over time are illustrated yearly to compare the results among different graft sources.

Full size image
Fig. 3: Five-year conditional net survival of children and older adults who have survived up to 5 years after transplantation.
figure 3

Five-year conditional net survivals of: (a) children and (b) older adults at year 0 to year 5 after the transplant. Changes in 5-year conditional net survival are compared between different graft sources for the age sub-groups using the methodology described in Fig. 2.

Full size image

AML

Considering the whole patient cohort, 5-year CS of PB was lower than BM from year 0 to year 5 and lower than CB from year 1 to year 3. Five-year CS of CB was lower than BM at year 0, but trends of CS for CB became similar to BM in the following years and were higher than PB from year 1 to year 3 (Fig. 2a and Table S2A). Trends of younger adults were similar to those of the all-age cohort. Five-year CS of PB was also lower than BM from year 0 to year 4 and lower than CB from year 1 to year 3. Five-year CS of CB was lower than BM at year 0 and higher than PB from year 1 to year 3 (Fig. 2b and Table S2B). For children, 5-year CS of PB with AML tended to be lower than BM only at year 0 (Fig. 3a and Table S2C). For older adults, 5-year CS of BM tended to be higher than PB at year 0 and year 5, and 5-year CS of PB was lower than CB only at year 4 (Fig. 3b and Table S2D).

ALL

For the whole patient cohort, 5-year CS of PB was lower than BM and CB from year 0 to year 2 (Fig. 2a and Table S2A). For younger adults, 5-year CS of PB tended to be lower than BM at year 0 and year 1, and was also lower than CB at year 1 (Fig. 2b and Table S2B). For children, 5-year CS of PB tended to be lower than BM at year 0 and year 1 and lower than CB only at year 0 (Fig. 3a and Table S2C). For older adults, trends were not different among different sources (Fig. 3b and Table S2D).

MDS

For all-ages cohort, 5-year CS of PB was lower than BM from year 0 to year 3, and CS of CB was lower than BM at year 0 (Fig. 2a and Table S2A). For younger adults, 5-year CS of CB was lower than BM only at year 0 (Fig. 2b and Table S2B). For older adults, 5-year CS of PB was lower than BM only at year 0 (Fig. 3b and Table S2D).

ML

For whole patient cohort, 5-year CS of PB and CB was lower than BM at year 0, and CS of PB was lower than CB at year 1 (Fig. 2a and Table S2A). For younger adults, 5-year CS of CB and PB were lower than BM only at year 0 (Fig. 2b and Table S2B). For children, 5-year CS of PB was lower than BM and CB at year 0 (Fig. 3a and Table S2B).

AA

For whole patient cohort, 5-year CS of PB was lower than BM from year 0 to year 5 and CS of CB was lower than BM only at year 0 (Fig. 2a and Table S2A). For younger adults, 5-year CS of PB was lower than BM from year 0 to year 2 and at year 4 and year 5, and CS of CB was lower than BM at year 0 (Fig. 2b and Table S2B). For children, 5-year CS of CB was lower than BM only at year 0 (Fig. 3a and Table S2C).

CML

For whole patient cohort, 5-year CS of CB was lower than BM only at year 0 (Fig. 2a and Table S2A). Sub analysis of CML was performed only for younger adults, and 5-year CS of CB was lower than BM at year 0 and year 1 (Fig. 2b and Table S2B).

MM

For the whole patient cohort, 5-year CS of CB was lower than BM and PB at year 0 (Fig. 2a and Table S2A). Sub analysis was performed for younger adults, and 5-year CS was not different among sources (Fig. 2b and Table S2B).

ATL

For whole patient cohort, 5-year CS of CB was lower than BM and PB at year 0 (Fig. 2a and Table S2A). As the number of patients were limited, sub analysis was performed only for younger and older adults. For younger adults, 5-year CS of CB was lower than BM and PB at year 0 (Fig. 2a and Table S2A). For older adults, 5-year CS of CB was lower than BM at year 0 (Fig. 2b and Table S2B).

Conditional OS using Kaplan–Meier method

Point estimates of 5-year conditional OS by diseases and age categories using Kaplan–Meier method are shown in Table S3. Survival trends were similar to those of conditional net survival. Point estimates of survival were comparatively lower than those of net survival.

Causes of death

Figure S2 summarizes the proportion of each cause of death category. Over time, the percentage of patients who died from secondary malignancy or post-transplant lymphoproliferative disorders, organ failure increased after transplantation. On the other hand, the contribution of relapse as a cause of death decreased after 3 years post-transplantation. The proportion of deaths classified as “cause of death unknown” increased in the later years after transplant.

Discussion

In this study, Japanese transplant registry data were analyzed to investigate long-term survival, utilizing the method used in population-based descriptive epidemiology studies. Five-year conditional net survival of children and younger adults with AA who had survived several years after the transplantation was nearly and exceeded 100% except for PB in younger adults, indicating that these patients had a similar chance of survival compared to the general population. On the other hand, 5-year conditional net survival of younger adults with MM receiving BM remained at 69% after 5 years survival, suggesting that some patients had recurrent disease. Furthermore, net survival of female patients was continuously higher than that of the male patients.

Differences in graft source continuously had an impact on 5-year CS through 5 years of transplantation for certain diseases. In previous studies, increased chronic GvHD in PB graft recipients with AA has been reported compared to BM among younger patients and adults [38, 39]. Increased chronic GvHD in PB could have continuously affect the survivals of the long-term survivors with AA. A previous study from the Center for International Blood and Marrow Transplant Research reported superior overall and GvHD, relapse-free survival of AML receiving BM grafts compared to PB in well-matched unrelated donor transplant [40]. The current study suggested extended effect of graft source on survival of the recipients with AML who had already survived for 5 years. Long-term outcomes of CB were favorable in certain indications such as younger and older adults with leukemia. Ethnic differences in CB outcomes, particularly in relation to GvHD, have been reported in a collaborative study between European and Japanese populations [41].

In this study, we used net survival to estimate conditional survival for two reasons. First, although the registry captures causes of death reported by the physicians, causes of death are sometimes reported as either unknown or others; moreover, some causes of death are not clear during follow-up [15]. Second, causes of death that are major long-term transplant complications, such as cancer and cardiovascular or respiratory infection, are also the main causes of death among Japanese or other general populations [42, 43]. These causes would be difficult to be categorized as the causes of or not directly related to the transplant. Since introduction of the method that are used in population-based descriptive epidemiology removes death other than that caused by the baseline disease or transplant-related causes, we aimed to investigate survival while excluding potential causes unrelated to the transplant or disease using this method.

We did not perform statistical hypothesis testing to compare the outcomes of different graft sources, as this study was exploratory research and to avoid the multiple testing problems. We analyzed the data by referring to the general population to see survival trends in different transplant sources by observing 95% CIs. Statistical analysis using observed survival rates with Kaplan–Meier method could be generally useful for comparison of factors related to survival within the transplant cohort. In this study, the trends of conditional survival within different grafts were comparable between conditional survivals using net-survival methods and Kaplan–Meier method (Tables S2 and S3).

Unlike prospective clinical studies, the transplant registry collects real world data; thus, the accuracy of data remains a limitation of the present study especially for assessing long-term survival of patients due to the loss of follow-up, which is a common limitation of using registry data. Second, this study included patients who underwent transplantation over 30 years ago. The impact of these patients on the outcomes remains uncertain, given the improvements in transplant outcomes in the past decade. Third, the analysis did not consider donor type for sub-analysis. Unrelated PB donors became available from the JMDP only after 2010, and the cumulative number of long-term survivors is still limited. Preliminary analysis suggests that conditional net survivals of related and unrelated BM recipients were comparable at later years after transplantation (result not shown).

In this study, we performed analysis of overall transplant patients, adapting methods used in population-based descriptive epidemiology. As few population-based studies have been performed in transplant settings [16], the results of this study can serve as a benchmark of survival for future studies.